Skip to main content

Full text of "Development and testing of INTRAS, a microscopic freeway simulation model : vol. 1, Program design, parameter calibration and freeway dynamics component development"

See other formats


Report No. FHWA/RD-80/106 






DEVELOPMENT AND TESTING OF INTRAS, 
A MICROSCOPIC FREEWAY SIMULATION MODEL 

Vol.1. Program Design, Parameter Calibration and 
Freeway Dynamics Component Development 

October 1980 
Final Report 




^O 1, 



Document is available to the public through 
the National Technical Information Service, 
Springfield, Virginia 22161 



Prepared for 

FEDERAL HIGHWAY ADMINISTRATION 
Offices of Research & Development 
Traffic Systems Division 
Washington, D.C. 20590 



FOREWORD 



This report presents the concepts and algorithms used in the INTRAS program 
which is a microscopic freeway simulation model which can be used to evaluate 
alternative designs of and control systems for urban freeways. It may also 
be used to study related subjects such as detector station spacing and the 
efficacy of incident detection algorithms. 

This report is the first volume in a four volume Final Report on the study, 
"Adaptation of a Freeway Simulation for Studying Incident Detecti. • and 
Control." 

This volume is being distributed by FHWA memorandum to interested 
researchers. A limited number of copies are available for official use 
from the Traffic Systems Division (HRS-31), Office of Research, Federal 
Highway Administration, Washington, D.C. 20590. Additional copies for 
the public can be obtained from the National Technical Information 
Service (NTIS) , U.S. Department of Commerce, 5285 Port Royal Road, 
Springfield, Virginia 22161. A small charge is imposed for copies 
provided by NTIS. 



>w^-^^--7V*^C 



Charles F. Scheffey 



NOTICE 



This document is disseminated under the sponsorship of the Department of 
Transportation in the interest of information exchange. The United States 
Government assumes no liability for its contents or use thereof. 

The contents of this report reflect the views of KLD & Associates. KLD 
is responsible for the facts and the accuracy of the data presented 
herein. The contents do not necessarily reflect the official views or 
policy of the Department of Transportation. This report does not constitute 
a standard, specification, or regulation. 












U\ 



m 

Technical Report Documentation) Pag» 



1. RtT>orV He 



FHWA/RD-80/106 



2. G 



o»tnv»w 



t Ace. 



No. 



4. Till* orxJ $<,blilli 



Development and Testing of INTRAS, a 
Microscopic Freeway Simulation Model 
Vol. 1. Program Design, Parameter Calibration and 
Freeway Dynamics Component Development 



7. AunW») 

D. A. Wicks, E. B. Lieberman 



3. Recipient's Cotolog No. 



5. Rroo't Don 



October 1980 



6. Performing Organization Coo* 



6. Performing Organi rolion Rapoti No. 



KLD TR-64 



T. Performing Organisation Hmp and Addroio 

KLD Associates, Inc. 

300 Broadway 

Hurrtington Station, NY 11746 



10. Wo.i, Unit No. (TRAI5) 

FCP 31C1-042 



II. Contract or Grant No. 

DOT-FH-11-8502 



12. Sponsoring Agoncy Now* ani AiWr«»» 

Office of Research and Development 
Federal Highway Administration 
U.S . Department of Transportation 
Washington, D. C. 20590 



13. Typo of Report ond Period Co»e'«d 

Final Report 



• 4. Sponsoring Ag»ncy Cod* 

T-0372 



IS. Suapl nl—y H»»i 

FHWA Contract Manager: S. L. Cohen (HRS-31) 



16. Aa>rro«t 

This series of volumes documents the work performed to adapt a freeway Simula^ 
tion model for studying freeway incidents. The resulting program, INTRAS 
(INtegrated TRA ffic Stimulation) , is a vehicle-specific time-stepping simula- 
tion designed to realistically represent traffic and traffic control in a 
freeway and surrounding surface street environment. 

This volume describes the detailed capabilities of INTRAS and its 
structural and procedural attributes. The calibration of traffic descriptive 
parameters are presented herein. The development of the simulation components 
which model the car-following lane changing and vehicle generation aspects 
of freeway traffic are also included. 

This volume is the first in a series. The others in the series are: 



Vol. No . 

2 
3 

4 



FIIWA No , 

80/107 
80/108 
80/109 



Short Title 

User's Manual 

Validation and Applications 

Program Documentation 



f\ 



d 






17. Kn*«J« Simulation, Freeway, Free- 
Simulation, Traffic Simulation, 
Freeway Control, Ramp Control, Free- 
way Operations, Incident Detection 



19. S^cunty Clt>5»i(. (ol rh.» f^o") 

Unclassified 



13. Distribution Statement 

No restrictions. This document is 
available to the public through the 
National Technical Information Service, 
Springfield, Va. 22161 



i0. S-*cufif-y Clo»)if. (of thi % poge) 

Unclassified 



21- Sy. of Poge* 

245 



22. P..C 



Fo™ DOT F 17C0.7 (9-72) 



Section 


1 




1. 


.1 


1. 


.2 


1. 


,3 


2 


' 


2. 


,1 


2. 


,2 


2. 


,3 


2. 


,4 


2. 


5 


2. 


,6 


2. 


7 


2. 


8 


2. 


9 


3 




3. 


1 


3. 


2 


3. 


3 


3. 


4 


4 




4. 


1 



TABLE OF CONTENTS 

Title Page 

INTRODUCTION 1 

Background 1 

Outline of Project Tasks 4 

Organization of Report 6 

CAPABILITIES OF THE INTRAS SIMULATION MODEL 7 

Network Definitions and Limitations 7 

Geometric Features of INTRAS 12 

Signal and Sign Control i6 

Traffic-Descriptive Features 17 

Incident Simulation Capability 18 

Surveillance System Simulation 19 

Freeway Traffic Responsive Control 19 

Detector Output Processing 24 

INTRAS User Options 25 

THE STRUCTURE AND METHODOLOGY OF INTRAS 28 

Functional Structure of INTRAS 29 

The User Interface 51 

INTRAS Storage Array Methodology 86 

INTRAS Error Procedures 98 

INTRAS FREEWAY PARAMETER CALIBRATION 100 

Vehicle Type Specific Calibration 100 
Parameters 



in 



TABLE OF CONTENTS (continued ) 

Section Title Page 

4.2 Other INTRAS Calibration Parameters 115 

5 LITERATURE REVIEW 124 

5.1 Analytical Car-Following Models 124 

5.2 Fail Safe Simulated Car Following 127 

6 SIMULATION DEVELOPMENT 127 

6.1 The Car Following Algorithm 127 

6.2 Lane Changing Development 139 

6.3 Vehicle Generation 145 

7 COMPONENT MODEL TESTING 14 6 

7.1 Calibration 146 

7.2 Validation 155 

8 SUMMARY 18 3 
APPENDIX A 185 
APPENDIX B 201 
APPENDIX C 208 
BIBLIOGRAPHY 220 
REFERENCES 228 



IV 



LIST OF FIGURES 
Number Title Page 

1 Sample Physical Freeway-Frontage 8 

Road Network 

2 Representation of Sample Network 9 

3 Typical Freeway Link Configurations 11 

4 Ramp-Freeway Interconnection Example 14 

5 Typical Metered Ramp Geometry 22 

6 Functional Structure and Information 

Transfer of INTRAS 31 

7 Sample of INPLOT Program Vehicle 

Trajectory Plot Design 80 

8 Sample of INPLOT Program Contour Plot 81 

Design 

9 Freeway Passenger Car Zero Grade 104 

Acceleration 

10 Non-Freeway Passenger Car Zero Grade 105 

Acceleration 

11 Speed-Distance Profiles for Intercity 110 

Buses and Heavy Single Unit Trucks 

12 Speed-Distance Profiles for Trailer 111 

Truck Vehicle Type 

13 Lane-Specific Mean Speed Ratio 118 

14 Cumulative Frequency of Speed to Mean 120 

Lane Speed Ratio 

15 Percent of Commercial Vehicle Population 123 

to the Left of Lateral Position L 



v 



LIST OF FIGURES (continued ) 
Number Title Page 

16 Platoon Behavior: PITT Algorithm-One 132 

Second Interval 

17 Platoon Behavior: PITT Algorithm-Three 13 3 

Second Interval 

18 Platoon Behavior: UTCS Algorithm-One 134 

Second Interval 

19 Platoon Behavior: UTCS Algorithm-Three 135 

Second Interval 

20 Platoon Behavior: Aerospace Algorithm- 136 

One Second Interval 

21 The Lane Changing Vehicles 141 

22 Driver Sensitivity (k) vs. Average 147 

Lane Capacity 

23 Intensity of Lane Changing as a Function 149 

of Volume for Two Lanes 

24 Intensity of Lane Changing as a Function 150 

of Volume for Three Lanes 

25 Intensity of Lane Changing as a Function 151 

of Volume for Four Lanes 

26 Courtesy Factor Calibration Capacity 154 

Measured at Ramp Nose 

27 Input Mean Spacing vs. Output Flow 156 

28 Lane Configuration for PINY Experiments 157 

2 & 6 

29 Speed-Volume Relationships 158 



VI 



LIST OF FIGURES (continued) 

Number Title Page 

30 Speed-Density Relationships 159 

31 Volume-Density Relationships 160 

32 Freeway Volume vs. Percent Trucks on 161 

Freeway 

33 Lane Volume vs. Percent Trucks on 162 

Freeway Lane 1 

34 Ramp Capacity as a Function of 163 

Acceleration Lane Length 

35 Comparison of Lane Changes - Lanes 171 

2-1, Experiment 2 

36 Comparison of Lane Changes - All 172 

Movements, Experiment 2 

37 Comparison of Lane Changes - Lanes 17 3 

1-8 , Experiment 6 

38 Comparison of Lane Changes - All 174 

Movements, Experiment 6 

39 Headway Distributions - Free and Con- 175 

gested Flow 

40 Comparison of Headway for Long Island 176 

Data - Total Volume 2 800 VPH 

41 Los Angeles Headways - Detector #56 178 

Shoulder Lane 

42 Los Angeles Headways - Detector #55 179 

Next to Shoulder Lane 

43 Los Angeles Headways - Detector #54 180 

Next to Median Lane 



vn 



LIST OF FIGURES (continued ) 

Number Title Page 

44 Los Angeles Headways - Detector #53 181 

Median Lane 

45 Ohio State Vehicle Trajectories Platoon 182 

of Twenty-Three Vehicles Showing 
Paths of Vehicle Numbers 1, 5, 10, 15, 
20 and 23 

46 INTRAS Supervisor Logic 186 

47 PORGIS Module Logic 187 

48 LIS Module Logic 190 

49 SIFT Supervisor Logic 193 

50 LOCON Suboverlay Logic 195 

51 HICON Suboverlay Logic 196 

52 POSPRO Module Logic 197 

53 INPLOT Module Logic 19 8 

54 INCES Module Logic 199 

55 SAM Module Logical Flow 200 






Vlll 



LIST OF TABLES 

Number Title Page 

1 Output of INTRAS Simulated Detectors 26 

2 INTRAS Routines 32 

3 PORGIS Routines 34 

4 PORGIS and LIS Subroutine Correspondance 37 

5 SIFT Routines 39 

6 SAM Routines 4 7 

7 INCES Routines 4 9 

8 INPLOT Routines 50 

9 INTRAS Data Categories 5 3 

10 Description of Input Stream 55 

11 Sample Freeway Link Definition Report 62 

12 Definition of Column Headings in 63 

"Freeway Link Definition" Report 

13 Sample Ramp and Surface Link Definition 64 

Report 

14 Definition of Column Headings in "Ramp 65 

and Surface Link Identification Report" 

15 Sample Sign and Signal Control Definition 66 

Report 

16 Definition of Column Headings in "Sign 67 

and Signal Control Definitions" Report 



IX 



LIST OF TABLES (continued ) 
Number Title Page 

17 Sample Entering Traffic Definition Report 68 

18 Definition cf Column Headings in "Entering 69 

Traffic Definition" Report 

19 Sample Surveillance System Definition 70 

Report 

20 Definition of Column Headings in 71 

"Surveillance System Definition" Report 

21 Sample Incident Definition Report 72 

22 Definition of Column Headings in 73 

"Incident Definition" Report 

23 Sample Freeway Statistical Report Design 74 

24 Definition of Column Headings in "Sample 75 

Freeway Statistical Report" 

25 Sample Ramp and Surface Statistical 76 

Report Design 

26 Definition of Column Headings in "Sample 77 

Ramp and Surface Statistical Report" 

27 Sample Freeway Station Headway and 78 

Speed Report Design 

28 Definition of Column Headings in "Sample 79 

Freeway Station Headway and Speed 
Report" 

29 Sample Point Processing Report Design 82 

30 Sample MOE Estimation Report Design 8 3 

31 Sample Incident Detection Report Design 84 

32 Sample SAM Data Element Paired Comparisons 87 



x 



LIST OF TABLES (continued ) 
Number Title Page 

33 Sample SAM Network Comparisons 88 

34 Sample SAM Link Specific Statistical 89 

Test Report 

35 Sample SAM Subinterval Specific Statis- 90 

tical Test Report 

36 Sample SAM Network Statistical Test 91 

Report 

37 INTRAS Calibration Normal Acceleration 106 

Rates for Low Performance Passenger 
Cars 

38 INTRAS Calibration Normal Acceleration 107 

Rates for High Performance Passenger 
Cars 

39 Multiplicative Factors Relating Passenger 109 

Car Acceleration on INTRAS Grades to 
Acceleration at 0% Grade 

40 INTRAS Calibration Normal Acceleration 112 

Rates for Buses and Heavy Single Unit 
Trucks 

41 INTRAS Calibration Normal Acceleration 113 

Rates for Trailer Trucks 

42 Ratio of Lane Speed to Mean Speed for 117 

Los Angeles Detector Data 

43 Ratio of Lane Volume to Overall per Lane 121 

Volume 

44 Ranking of Car-Following Algorithm 140 

45 Mean Frequency of Lane Changing vs. 152 

Probability of Lane Changing and Volume 



XI 



LIST OF TABLES (continued) 
Number Title Page 

46 Comparison of Simulation Outputs and 166 

Field Data for Experiment 6 

47 Comparison of Simulation Outputs and 167 

Field Data for Experiment 6 

48 Correlation Analysis of PINY Data and 168 

Simulation Outputs 

49 Comparison of Simulation with Given Data 170 



xn 



1. INTRODUCTION 

A program of major emphasis has been launched by the 
Federal Highway Administration (FHWA) to develop and im- 
plement incident detection strategies and to integrate 
these with surveillance and control policies to alleviate 
traffic congestion on the nation's freeways. A major ele- 
ment of this program is the development of a microscopic 
simulation model which can be utilized as a tool for evalu- 
ating alternate candidate solutions to this problem. Con- 
tract DOT-FH-11-8502 calls for the research agency to de- 
sign, program, calibrate, validate, and demonstrate such a 
computer simulation model. 

Volume I of this Final Report describes the work 
effort for Tasks A and B: 

• Adaptation of the UTCS-1 Network Simulation 
Model Logic , 

• Validate Simulation Components. 

Volume II is a User's Manual for the resulting traffic 
simulation program, INTRAS . Program validation activities 
and the application of the validated model are described 
in Volume III. Detailed program documentation constitutes 
Volume IV. 

1 . 1 Background 

The two major subject areas which must be syn- 
thesized in this project are: 

• Freeway Operations, Surveillance, Control 
and Incident Detection; 

• Simulation of Traffic. 

Brief overviews will be presented of each subject area. 

1.1.1 Freeway Operations and Incident Detection 

In recent years, increased attention has 
been focused on the need for developing effective freeway 
incident detection techniques. Moskowitz , in his review of 
research needs in traffic surveillance (Ref.l) stated that he 
believed the single most important problem in urban free- 
way traffic operations is the detection of stopped vehicles 
and the necessary steps required to remove the stoppage. 



Flow-disruptive incidents are a substantial cause of 
congestion and considered by some to be even more of a 
problem than recurring daily congestion. In the Los 
Angeles area the California Division of Highways estimates 
that at least one-half of the delays to traffic can be 
attributed to incidents occurring along the freeway. 
This finding supports the earlier work done in Chicago by 
the Expressway Surveillance Project which also estimated 
the delay from incidents as at least one-half of the total 
delay to traffic. 

Incidents occur predictably in frequency, but 
unpredictably by time and location. Often appearing 
during peak periods, these incidents severely reduce the 
level of service on the urban freeway. In most cases 
they significantly reduce the capacity of the freeway 
although in some cases the traffic demand may still be 
lower than the reduced capacity. Incidents which cause 
capacity reductions are a problem for demand volume either 
above or below capacity levels. Under conditions of light 
traffic demand the impact of the incident is to primarily 
slow traffic past the incident. Crane (Ref. 2) estimated 
that there are 6400 accident-related lane blockages each 
year on the 50-mile Detroit freeway network. These 
accidents only account for 7% of all lane blockages. He 
concludes that 14% of the time there are one or more 
lanes blocked by an incident in the system. The average 
time, that a traffic lane is blocked by an incident tends 
to be short, in the range of four to six minutes. 

The objective in managing traffic operations at the 
scene of an incident is to minimize the hazard and delay 
caused to both the motorist involved and those passing 
through the affected area. The treatment of an incident 
is considered in four steps: 

a) Detection of the incident 

b) Identification of location and type of incident 

c) Service response to the incident 

d) Restoration of traffic operations. 

The first two steps are addressed by this project. 



1.1.2 Simulation of Traffic 

Digital computer programs to simulate 
traffic flow have been developed over a period of three 
decades. The great appeal of the simulation approach is 
that this technique offers the user an opportunity to 
evaluate alternative strategies before implementing them 
in the field. Thus the optimal strategy may be identified 
prior to the commitment of substantial funds for implemen- 
tation of large systems. 

Despite the intense interest in this field, it is only 
recently that a very limited number of such models have 
been applied productively to solve real-world problems in 
traffic engineering. The reasons for the limited number 
of successful models can be traced to the complexities of 
the physical problem of traffic dynamics and the resulting 
sophistication of program logic required to replicate 
these real-world events properly. While these difficulties 
apply tothe smallest traffic element — the isolated inter- 
section — they are greatly amplified when the scope of the 
problem is enlarged to encompass a network representation 
of a roadway system. 

Simulation models may be classified as microscopic or 
macroscopic. The term "microscopic," as used here, denotes 
a model which simulates the movements of individual 
vehicles. "Macroscopic" denotes a grouping of vehicles and 
the application of flow relationships to determine success- 
ive traffic states. 

• A microscopic model generally requires a larger 
programming and debugging effort, exhibits more 
stringent storage requirements and consumes more 
computing time, while providing greater resolution 
and potentially more accuracy, relative to the 
alternative. 

• A macroscopic model, on the other hand, while 
being more economical in every way, may be unable 
to describe a complex process adequately, yielding 
inaccurate or misleading results which are wholly 
unacceptable . 



It is seen that a careful tradeoff between level of de- 
tail and economy of operat .jn is an essential ingredient in 
a successful, large-scale .-simulation model. Clearly, the 
degree of simulation microscopy must be tailored to the na- 
ture of the process it is describing. This requirement im- 
plies that the designer must possess a keen understanding 
both of the dynamics of the process and of efficient simu- 
lation methodology. 

1. 2 Outline of Project Tasks 

The project defined by the subject contract is 
segmented into the following tasks: 

Task A - Adapt the UTCS-1 Network Simulation Model for 
Freeway Applications 

The UTCS-1 simulation model (Ref. 3,4) has been de- 
veloped under FHWA sponsorship over the last five years 
as a vehicle for evaluating urban traffic systems. This 
program is a stochastic, microscopic simulation model that 
traces the trajectory of each vehicle along the streets 
(links) of a specified network. The control devices 
accommodated included pre-timed and actuated traffic sig- 
nals, stop signs and yield signs. Other features include 
surveillance systems (detector deployments) , bus traffic 
and random blockages (events) . 

The simulation procedures of the UTCS-1 model are to 
be adapted to the simulation of traffic on a freeway and 
parallel service facility. The resulting model, INTRAS, 
(Integrated TRAffic Simulation) , will retain the stochas- 
tic microscopic nature and traffic operation and control 
simulation capabilities of UTCS-1. The scope of the pro- 
gram will be expanded and will include specially designed, 
calibrated and validated component models representing the 
dynamics of freeway flow. 

Task B - Validate Candidate Components 

The candidate component models, representing freeway 
traffic dynamics, will be subjected to statistical tests to 
ascertain their validity. Comparable field data bases will 
be used to represent the "real" freeway environment. Output 
of the component models, for demand levels present in the 
field data bases, will be compared with the corresponding 
field statistics. 



Task C - Program the Simulation Model 

The INTRAS simulation model will be programmed so as to 
run effectively in a 24 OK byte partition of the FHWA IBM 
360/65 computer. The model will be thoroughly debugged by 
exercising all logical paths. 

Task D - Validate and Refine the Simulation Model 

The debugged simulation model will be exercised, over a 
wide range of traffic demand, to determine whether the 
integrated components combine to produce a valid overall 
model. Comparisons of a simulation output with field data 
will be performed for at least the following traffic mea- 
sures of effectiveness (MOE) : 

• Total travel time and vehicle miles for each 
subinterval (period of constant demand volume) 
and roadway section (subsystem) 

• Spot speed and headway distribution at one point 
in each subsystem for each subinterval 

• Frequency of lane changes in each subsystem 
for each subinterval 

• The lane distribution of traffic at selected 
points for each subinterval to determine the 
effect of distance from on-ramps and off-ramps 

If the model proves to be invalid, it will be appropriately 
modified and retested until validation can be demonstrated. 

Task E - Validate Incident Detection Algorithms 

Incident detection algorithms supplied by FHWA will be 
programmed as an integral part of the INTRAS model. These 
algorithms will act on output of the INTRAS simulation to 
detect simulated incidents. For each incident detection 
algorithm, situations from a range of traffic conditions will 
be simulated and, for each condition, the following three 
parameters will be calculated for both the beginning and 
ending of incidents: 



• Percent successful detection 

• Percent false alarm 

• Distribution of time to detection. 

The criteria of Task D and existing traffic stream 
data will be used to determine the validity of the detec- 
tion algorithms within the structure of the INTRAS simula- 
tion model. 

Task F - Apply the Simulation Model 

Two simulation application studies will be performed. 
First, each of the FHWA-supplied incident detection algor- 
ithms will be applied in a parametric evaluation. The 
three parameters of Task E will be evaluated over a range 
of two independent variables: volume and detector spacing. 

Second, the most promising incident detection algorithm 
and detector configuration will be applied in conjunction 
with a ramp control policy. The ability for the ramp 
control policy to ameliorate the disturbance caused by the 
incident will be evaluated for a range of volumes. Incident 
removal will also be studied. 

1 . 3 Organization of Report 

The subject of Sections 2 through 4 of this Final 
Report consists of the design, development and component 
calibration and validation activities accomplished prior to 
the actual simulation model programming (Task C) . These 
sections describe the computer program design and the cali- 
bration of parameters and arrays, needed by the model to 
represent various traffic characteristics. Sections 5 
through 7 describe the development, calibration and valida- 
tion of the various components which model the dynamics of 
freeway traffic (car-following, 1 ane -changing , and vehicle- 
generation). The conclusion, Section 8, summarizes the com- 
puter program design activities. 



2. CAPABILITIES OF THE INTRAS SIMULATION MODEL 

Those design characteristics which define the INTRAS 
model capabilities are described in this section. Specific 
features required by the Task A work statement, as well as 
others which augment the performance or usefulness of the 
simulation, are included. 

2 . 1 Network Definitions and Limitations 

The geometric representation of a roadway system in 
the INTRAS simulation model is accomplished by constructing 
a network analog of links (roadway segments) and nodes 
(intersections or geometric discontinuities) . Figures 1 
and 2 illustrate a typical road system and its network 
representation, respectively. To realize economy of stor- 
age, and to permit appropriate logical treatment for road- 
way sections of diverse characteristics, three link types 
are defined for INTRAS. 

A Surface link is defined as a roadway segment ser- 
vicing one direction of traffic. The nodes at each end 
represent grade intersections. Each "surface" link extends 
from the upstream stopline to the downstream stopline as 
in the following sketch: 



Surface Link, (j,i) 



1+ 



Surface Link (i,j) 



Intersection 
(node) i 



Intersection 
(node) j 



As indicated (in the sketch) a link is normally identi- 
fied by the upstream and downstream node numbers. Each 
"surface" link may consist of up to 5 lanes in width. Two 
of these lanes may be turning pockets (one left and one 
right) which do not extend for the full link length. 




QJ 




Cn 




fO 




■P 




c 









u 




fa 




>i 




rfl 




5 









CD 




H 




fa 




rH 




IC 




u 




-H 


-V 


CO 


u 


> 


o 


xi 


5 


ft 


+J 




0) 





-z- 


rH 




o^n 


B 


H3 


(0 





CO 


ex 




u 

2 

•H 
fa 




M 
U 
O 

+J 

CD 
Z 



e 

fO 

in 
o 

c 
o 

•H 
-P 

fd 
-p 
c 
a) 
w 

CD 

a 

CD 



cm 

CD 

P 

-H 



to ' 


1 


s 


>i 


CD 


n 


CD 


-p 


U 


c 


fo 


H 



Vehicles traversing an INTRAS "surface" link will be 
moved at one-second intervals utilizing the logic established 
in the UTCS-1 urban traffic simulation model. This resolution 
properly replicates (Ref. 4) the dynamics of traffic on 
urban networks. 

A Freeway link is defined as a one-way roadway segment, 
of a controlled-access highway, characterized by generally 
constant geometric characteristics (grade, curvature, number 
of through lanes) . The extremities of a "freeway" link 
correspond to either ramp junctions or significant geometric 
changes. Each "freeway" link may contain up to five through 
lanes and two auxiliary lanes. Each auxiliary lane may be 
described as "acceleration", "deceleration" or "full", as 
defined below: 

Auxiliary 

Lane Type Definition 

Acceleration A lane which extends from the 

upstream extremity of a free- 
way link to some mid-link 
position 

Deceleration A lane which extends from a 

mid-link position to the 
downstream extremity of a 
freeway link 

Full A lane which extends for the 

full length of a freeway link 
with at least one end con- 
necting to an on or off-ramp 

Auxiliary lanes may occur on either the left or right- 
hand side of the roadway. Typical "freeway" links are 
illustrated in Figure 3. 

Vehicles traversing "freeway" links move in accordance 
with the logic of component models specially designed for 
INTRAS (see Sections 5-7). 



10 



Change 
in Grade 
Node j 



Ramp 

Junction 
Node i 



J-- A _' 



Freeway Link (i,j) 



Ramp 

Junction 
Node i 



Node j 




Node i 




■Freeway Link (i,j 



Change in 

Curvature 

Node i 



_ _D V 



Freeway Link (i,j)- 



Node j 




Node 


> j Nod 


pi 


F 












-* — 




< — Freeway Link ( i , i )■*■ 





Node i stS 






•Freewav Link ( i , i >> 



/ 



s 




'4 



Node 


j Node 


3 




/ ft 






-* 






-< 












r^ jir 


D 7 




■♦-Freewav Link (i,j)~*- 





Key : 

A Acceleration Lane 
D Deceleration Lane 
F Full Auxiliary Lane 

Figure 3: Typical Freeway Link Configurations 



11 



A Ramp link is defined as a one-way non-freeway roadway 
segment which connects directly to a freeway link. Ramps 
may be one or two lanes in width. "Ramp" links are further 
characterized as either on or off -ramps indicating that end 
of the link that connects to the freeway. 

The same logic is applied to move vehicles on "ramp" 
links as for "surface" links. 

Absolute restrictions on the network size and traffic 
density, which may be accommodated by INTRAS , exist by 
virtue of the node, link and vehicle identification 
structure. Thus, links are identified by a 3-digit number 
(within link type) so that there may be no more than 999 
links of each type. 

Similarly, the maximum number of vehicles permitted on 
each link type (freeway, ramp, surface), is 9999. Node 
numbers in the interior of the network may take on values from 
1 through 699. Nodes which describe the outer extremities 
of entry and exit links are limited to values from 700 ■*■ 799 
and 800 ->- 899, for freeway and surface extremities, 
respectively. There is an absolute link length limitation 
of 3265 feet for surface and ramp links and 9800 feet for 
freeway links. None of the above limits should pose any 
significant problem to the user. 

Relative restrictions on network size and traffic 
volume exist and vary directly with the available computer 
memory. Dynamic allocation of storage to link and vehicle 
arrays (Section 3) is performed, internal to the INTRAS 
program, to assure full utilization of the available memory. 

2 . 2 Geometric Features of INTRAS 

To model a roadway system in sufficient detail to 
replicate "real world" traffic statistics, it is necessary 
to accommodate those geometric features which significantly 
affect traffic performance. Geometric features (lanes, 
lengths, number of links) which represent numeric limits on 
the size of roadway systems which may be simulated were 
treated in the previous section. Other significant geometric 
features of the INTRAS design are described below. 



12 



Intersections - The junction of surface links with 
either other surface links or ramp links are modeled as 
in the UTCS-1 program. Each intersection is identified by 
a unique node number. Links are identified by the ordered 
pair of node numbers which identify their upstream and 
downstream extremities. There may be up to four links 
approaching, and four links departing, a given intersection 
(node) . 

Vehicles on each approach link to an intersection may 
have up to three destinations (receiving links) upon pass- 
ing through that intersection. Each of these receiving 
links is entered by performing the associated traffic man- 
euvers: left turns, through movement or right turns. Left 
turners will seek gaps in opposing traffic; right turners 
will slow before turning, etc. 

Freew ay-Freeway and Freeway-Ramp Inter conne ctions - The 
lane alignment of freeway links and on-ramp links with the 
next downstream freeway link is defined by two input speci- 
fications. First, the number and type (through, auxiliary) 
of lanes which comprise each link is specified. Second, 
the lane in the downstream link which receives traffic from 
the right-most through lane of the upstream link must be 
identified . 

Lanes are labeled for identification via the following 
convention. The through lanes of each freeway link (and 
all lanes of ramp links) are numbered sequentially from 
right to left (i.e., the right-most through lane is always 
labeled "1"). The left-hand auxiliary lanes are numbered 
6 and 7, respectively, with lane 6 adjacent to the freeway 
lane. Right-hand auxiliary lanes are numbered 8 and 9, 
respectively, with lane 8 adjacent to the freeway lane. 

Freeway links are logically connected to downstream off- 
ramps by specifying the number of ramp lanes and whether 
it is a right-hand or left-hand off-ramp. The outside 
lanes on the designated side of the freeway are then inter- 
nally assigned as connecting to the off-ramp. 

As clarification of this convention, the following ex- 
ample is provided: 

Figure 4a illustrates a roadway section containing four 

13 



\ 




I I 



00 

\ 



00 



CN 



eg 




CO 



CO 



Cx] 



Cxi 



^r 






rti 




CD 
iH 

Cu 

e 

X 

w 

c 
o 

•H 

-p 
o 

<D 
C 

c 
o 
u 

CD 
■P 
C 

H 

>i 

<D 

CD 
5-1 

I 

a 

e 

cm 



CD 

u 
tr 

■H 

Cm 



14 



freeway links, one on-ramp, and two off -ramps. Figure 4b 
contains the network idealization for the same roadway 
section. In this example, the alignment of freeway links 
and on-ramps is given in the following table: 

Feeding Receiving Receiving Lane 

Link Link for Lane 1 Traffic 

1.2 2,3 1 

2.3 3,4 8 

3.4 4,5 1 

6,2 (on-ramp) 2,3 8 

The off-ramp alignment is implicitly determined in 
INTRAS by the specification that left exiting vehicles from 
2,3 enter 3,7 and right exiting vehicles from 3,4 enter 4,8, 
This specification, coupled with the number and definition 
of lanes in the four links, is sufficient to define align- 
ment. 

Grade Specification INTRAS has been designed to ac- 
cept link-specific grade as input. Thus, it is proper to 
define a continuous section of roadway (containing a sig- 
nificant change in gradient) as two contiguous links, with 
a node defined at the point where the grade changes. The 
INTRAS logic examines the link-specific grade specification 
to modify several operating parameters (see Section 4). 

Curvature As for grade, a change in horizontal curva- 
ture is sufficient reason to segment a roadway section into 
two links. Two methods of limiting vehicle performance on 
horizontal curves are available in the INTRAS design. 
First, a lowered value of desired free-flow speed may be 
defined for an affected link. Although easy to apply, this 
method presumes some pre-analysis on the part of the user. 

Second, radius of curvature, super elevation and pave- 
ment condition may be defined. An internal table is re- 
ferenced to determine friction coefficient from pavement 
condition. The basic equation for vehicle operation on a 
curve (Ref. 5), is then used to generate an upper bound for 
desired free-flow. 

15 



V = J 15R(e+f) 

where, e = rate of roadway superelevation, foot per foot 

f = Friction coefficient for given pavement condition 

R = Radius of curve in feet 

V = Vehicle speed, miles per hour. 

The simulation model applies the minimum of the input free- 
flow speed, and the curvature dictated upper bound, to 
traffic on the subject link. 

Lane Separation - The typical freeway often contains 
sections where lane changing is physically prohibited by 
virtue of barrier curbs or traffic islands. These restric- 
tions are designed to segregate through traffic from 
weaving traffic, or, to guide vehicles around some obstruc- 
tion (bridge abutments, etc.). INTRAS is designed to ac- 
cept physical barriers of this nature on a link-specific 
basis. 

2 . 3 Signal and Sign Control 

Each intersection in a simulated study network re- 
quires a control policy to establish the right-of-way for 
approaching vehicles. Similar to the UTCS-1 program, INTRAS 
has the ability to simulate both fixed-time signal control 
and sign control. Provision has been made for the modular 
inclusion and referencing of specially coded subroutines to 
model traffic responsive signal control. 

Fixed Time Signal Control - Intersections of an INTRAS 
simulated network may be controlled by fixed time signals of 
up to six control intervals each. During each interval, one 
of the following standard signal configurations is applied 
to control each of the approach links: 

Amber 

Green 

Red 

Red with Green Right Arrow 

Red with Green Left Arrow 

No Turns - Green Through Arrow 

Red with Left and Right Green Arrows 

No Left Turn - Green Through and Right 

The duration of each control interval is user-specified. 

16 



Sign Control - Each intersection not controlled by a 
fixed-time signal is controlled by either stop or yield 
signs. The user must specify which approaches face such 
signs. For the common situation, where no control of any 
kind is present, the INTRAS user will need to specify 
yield signs for one approach direction to indicate the 
minor street. 

Ramp Metering - Four algorithm modules have been coded 
for controlling vehicles entering a freeway from on-ramps 
(see Section 2.7.1). 

Freeway Traffic Diversion - Two algorithm modules have 
been coded for diverting freeway vehicles to alternate 
routes (see Section 2.7.2). 

Traffic Responsive Intersection Control - A control 
module containing procedures similar to those in the 
UTCS-1 simulation have been implemented conforming to the 
NEMA standard. 

An additional control has been designed for freeway 
link application. Advisory Signs will be simulated which 
inform freeway drivers of the presence of a downstream 
exit. Vehicles passing the acquisition point for these 
signs will alter their desired lane if they are assigned 
to exit the freeway at the indicated ramp. 

2. 4 Traffic Descriptive Features 

Each driver-vehicle pair in a traffic stream be- 
haves as an individual entity having different motivations 
and standards of performance from those around it. This 
quality must be modelled in INTRAS to achieve the proper 
stochastic variation in individual vehicle performance. 
To accomplish this, the INTRAS design provides for five 
vehicle types, each possessing its own family of vehicle 
characteristics (length, speed, acceleration profile, etc.! 
These characteristics may be revised as an option, so that 
the particular vehicle types chosen for the basic INTRAS 
model do not constitute a limitation on the user. The 
vehicle types chosen for the initial development of INTRAS 
are : 

High Performance Passenger Car 

Low Performance Passenger Car 

Intercity Buses 

Single Unit Trucks 

Trailer Truck Combinations- 

17 



Further discussion of vehicle-type specific characteris- 
tics appears in Sections 3.1.9 and 4. 

Variations within vehicle type are attributed to 
differences in driver performance. Decile distributions of 
these characteristics (variation about mean free-flow speed, 
queue discharge headway, etc.) are implemented in the 
INTRAS model as in the UTCS-1 program. 

The two most important elements describing the traffic 
assignment on a given network are entering volume and rout- 
ing. The INTRAS design allows specification of entering 
volumes, by vehicle type. The volume for each entry link 
is held constant over a period of simulated time referred 
to as a subinterval. At the end of each subinterval any 
number of these entering demand volumes may be revised. The 
duration of each subinterval is a user specification, there- 
by providing complete freedom in the variation of traffic 
loading with time. 

Routing is normally performed by specifying the per- 
centage (or count) of vehicles negotiating each possible 
turn movement on a link-specific basis. Turn movements may 
also be varied by the user from subinterval to subinterval. 
See Section 2.7 for a discussion of other routing techniques. 

2 . 5 Incident Simulation Capability 

A comprehensive freeway incident simulation pro- 
cedure has been designed for INTRAS. The user may specify 
either blockages or "rubbernecking" to occur on a lane- 
specific basis. Each incident may occur at any longitudi- 
nal position on a freeway link and extend for any desired 
length of time. 

The character of an incident may be changed with time. 
It is possible to specify, for example, a two-lane blockage 
which, after some specified duration, becomes only a one- 
lane blockage. The lane from which the blockage is removed 
may then become unrestricted or subject to "rubbernecking". 

"Rubbernecking" may be applied, without a corresponding 
blockage, to simulate a shoulder incident. The user will 
input a factor indicating the percentage reduction in speed 
for vehicles traversing the affected lane segment. 

18 



2 . 6 Surveillance System Simulation 

To render the INTRAS model an effective tool for 
evaluating a "real world" traffic performance evaluation 
and control techniques, it is necessary to simulate "real 
world" information gathering (surveillance) systems. Three 
types of traffic detectors are to be simulated by INTRAS. 

Doppler radar detectors are characterized in the model 
by the lane and longitudinal location at which vehicles are 
detected. Each simulated vehicle crossing this location 
will cause the surveillance logic to output speed and time 
of actuation. 

Short inductance loops are characterized by lane and 
longitudinal position, and loop length. Either of two 
output methods may be chosen by the user. A digital output 
form may be selected to simulate the time interval scanning 
method prevalent in many control systems . The output for 
this method consists of an (occupied, not occupied) status 
indication. Normally the detector scanning interval will 
be of shorter duration than the simulation time step. The 
on, off status of each detector for each scanning interval 
is obtained by interpolating vehicle position across the 
simulation time step, assuming constant acceleration. 

An analog output form may be chosen which outputs the 
time and duration of actuation for each vehicle. 

Coupled short inductance loops are described to INTRAS 
as two single short loops of equal length. The downstream 
loop is located by lane and longitudinal position. The up- 
stream loop is implicitly located by defining the separa- 
tion distance between the pair. The output for coupled 
pairs may either be analog or digital as for single loops. 

All surveillance system output is made available for 
subsequent treatment by point processing (detector specific) 
and system evaluation (measure of effectiveness — MOE) esti- 
mation and incident detection modules. Internal arrays are 
maintained to service the dynamic control logic (see the 
following section) . 

2 . 7 Freeway Traffic Responsive Control 

Vehicles entering the freeway via on-ramps may be 

19 



subjected (as a user option) to a variety of control 
techniques. In parallel to, or independent of on-ramp 
control, diversion of freeway vehicles to a parallel ser- 
vice facility may be simulated. Both metering of ramp ve- 
hicles and diversion of freeway traffic are accomplished on 
a node-specific basis via specially coded subroutines. 
This subroutine structure permits a modular replacement of 
control policies. Control parameters are entered through 
the normal input stream for each affected node, in a gen- 
eral format acceptable to all control policies. 

2.7.1 On-Ramp Controls 

Four methods of on-ramp control are to be 
implemented in the INTRAS model. A typical geometric con- 
figuration of a metered on-ramp site is shown in Figure 5. 
The ramp signal (at C in the figure) is assumed to be up- 
stream of the ramp-freeway interface at B. The downstream 
section of the physical ramp, link (C,B) is represented as 
a ramp link . The upstream portion (D,C) is represented as 
a surface link , subject to the normal queue discharge logic 
applied at all signalzied intersections. Optionally, node 
C may be removed, and the ramp control may be applied at 
node D. In this event (D,B) is represented as ramp link . 
The signal indication at D then specifies the ramp metering 
rate, as well as the control to through traffic. 

Clock Time Metering - To simulate clock-time control of 
on-ramps, one fixed metering rate (vehicles per minute) is 
specified at each such node. A count-down clock is assign- 
ed to each associated on-ramp and the signal is set to 
"green" each time the clock returns to zero. The signal is 
maintained at "green" until a vehicle is discharged, and is 
then set to "red". A non-compliance percentage (user spe- 
dified) is applied to vehicles arriving during the "red" 
signal. The indicated percentage of vehicles will be dis- 
charged through the "red" signal. 

Demand/Capacity Metering - An evaluation of current 
excess capacity, immediately downstream of the metered on- 
ramp, is performed at user specified intervals. A maximum 
metering rate is calculated such that capacity of this 
freeway section is not violated. This metering rate is 



20 



applied as for clock-time metering. A minimum metering rate 
of 1 vehicle/20 seconds is applied to ensure that waiting 
vehicles are not "trapped" (i.e., as in Section D, C of 
Figure 5) . 

In addition to the evaluation period, the user must 
specify the following parameters for each on-ramp operating 
under demand/capacity metering control: 

• Capacity at downstream freeway station 

• Freeway link to which capacity applies 

• Identify detectors on that link which 
provide input to the metering algorithms 

Speed Control Metering - The procedure for this form of 
ramp metering is rather similar to the demand/capacity descrip- 
tion above. A freeway link detector station must be 
established and identified at which speed evaluations are 
to be made to establish a metering rate. Generally, this 
location will be upstream of the on-ramp, although the 
logic will not preclude other placements. The user must 
specify a table of speeds, and metering rates, for each 
speed controlled on-ramp. As each evaluation period con- 
cludes, the prevalent speed, at the datum freeway station, 
will be compared to the tabular speed minimums to 
determine the proper metering rate. 

Gap Acceptance Merge Control - This method of ramp 
control employs the ramp signal control to release ramp 
vehicles so as to merge smoothly into gaps detected in the 
outside freeway lane traffic. The input required to imple- 
ment this procedure is simplified in that no evaluation 
period nor metering rate criteria is required. A coupled 
pair of loop detectors must have been defined (via 
surveillance system specifications) in the outside lane of 
the upstream freeway link. The link identification and 
detector position must be identified to the gap acceptance 
algorithm, for each ramp, as well as the minimum acceptable 
gap size. 



Gaps are expressed in units of time. 



21 




FRONTAGE ROAD 




E 



KEY: 



Auxiliary Lane of 
Freeway Link B, A 



Left Turn Pocket of 
Surface Link E, D 






Figure 5: Typical Metered Ramp Geometry 



22 



Gaps detected in the traffic stream are projected down- 
stream to the merge point. The merge point gap size will be 
adjusted to reflect the relative speeds of the leading and 
following vehicles. As in the design of a physical system, 
the user will have to excercise care that one of the follow- 
ing two logical absurdities does not occur: 

• The detector is so close to the merge point that 
vehicles at the ramp signal can not be released 
in time to enter an acceptable gap 

• The detector is so far upstream as to significantly 
affect the accuracy of the projected gap size 

All of the above ramp control methods are subjected to 
built-in distributions reflecting (by driver-type) , maximum 
tolerable queue delay for queue leader, and willingness to 
join a queue. These distributions are imbedded in the 
modular ramp control subroutines. 

2.7.2 On-Freeway Diversion 

Two procedures for diverting freeway 
vehicles to a parallel service facility will be programmed 
as subroutines of INTRAS. Both procedures will involve 
assignment of some portion of the through vehicles (at each 
freeway-off ramp junction) to the off-ramp. Each diverted 
vehicle will be assigned to a user specified path, eventual- 
ly leading either out of the network or back to the freeway. 

Clock Time Diversion. This diversion method will apply 
a user specified percent of through vehicles to be diverted 
to each affected off-ramp. The time at which diversion 
begins must also be specified as well as the subsequent 
routing for each vehicle. 

Least Time Path Diversion . For this procedure, travel 
time will be monitored for freeway paths and user-specified 
alternate paths which rejoin the freeway at some downstream 
node. When, and if, an all-freeway path is more time 
consuming than its alternate, vehicles will be diverted at 
the appropriate off -ramp. The number of vehicles diverted 
will be calculated such that the total ramp loading does 
not exceed ramp capacity. 



23 



2 . 8 Detector Outpu t Processing 

The simulated surveillance system produces output 
analagous to that generated by an on-line system. This 
output is stored on a peripheral device as a sequential file. 
At the conclusion of the simulation, this file may be saved 
for subsequent processing or analyzed immediately. The 
following procedures for analyzing detector data are includ- 
ed in the INTRAS design. 

• Point p rocessin g procedures process each 
individual detector's output to generate 
local estimates of traffic flow parameters 

• Measure of Effectiveness (MOE ) estimation 
procedures, provided by FHWA, generate 
system wide or subsystem (link) -specific 
parameters 

• Incident detection procedures, provided 
by FHWA, analyze the detector data to 
identify the occurence of capacity reducing 
incidents. 

All evaluations aggregate data over time intervals 
of user specified duration. 

The following subsections describe the functions of 
these procedures in more detail. 

2.8.1 Point Processing 

Each detector on a roadway emits basic 
information as "raw" data. This "raw" data takes on 
different characteristics depending on the nature of the 
detector and the communications method. In "digital" mode 
each detector is polled at fixed intervals to determine 
current status (on-off, occupied-not occupied, etc.). In 
analog mode each detector sends a signal whose amplitude 
is proportional to the current measurement. INTRAS may 
operate in either digital or analog mode in simulating loop 
detectors. Simulated doppler radar detectors only operate 
as analog. 



24 



Table 1 describes the output of the detector types. 
Also shown are the parameters evaluated via point pro- 
cessing procedures. Each evaluation is detector specific. 
An assumed value of vehicle length is embedded in the point 
processing procedures to permit calculation of speed for 
single loops. 

2.8.2 MOE Estimation 

Procedures specified by FHWA will be imple- 
mented in INTRAS to estimate traffic performance parameters 
for freeway sections between detector stations. These 
procedures will operate on the data generated by detectors 
at the upstream and downstream stations. Estimates of 
area-related parameters (travel time, density, delay, etc.) 
will be generated to characterize traffic performance in 
the study sections. 

2.8.3 Incident Detection 

The primary goal of the subject project is 
to develop a simulation model to be used for the study of 
freeway incident detection and control. As described 
earlier, the INTRAS design includes comprehensive incident 
and surveillance system simulation capabilities. The 
INTRAS model will contain algorithms (to be provided by 
FHWA) to analyze the detector data and determine whether 
or not an incident has occurred on the simulated freeway. 
Either "raw" detector data or the results of the point pro- 
cessing procedures may be used as input to the incident 
detection algorithms. 

The output of these procedures will be the time of 
detection of both the onset and end of each incident. When 
all detector data for the full simulation run has been 
processed, a comparison of the performance of the detection 
algorithms with the actual simulated history will be gener- 
ated. The MOE of interest will include: time till detec- 
tion (both onset and end of incident) , percent of real 
incidents detected, and percent false alarms. 

2.9 INTRAS User Options 

Numerous peripheral functions are designed into 
the INTRAS model to enhance its usefulness in the study of 



25 



w 
u 
o 

•P 

u 

<D 
-P 

(D 
Q 

'd 
d) 

-P 
fO 

.H 

p 
g 

•H 



Eh 
S3 



O 

4-> 

P 

-P 
P 
O 




■H 

(0 
Eh 



-P 
P 

c 
-p 
p 

o 

c 

•H 

w 
w 

Q) 
U 
O 

(^ 

-P 

c 

•H 

o 



•P 

P 

-P 

P 

o 



-P 
•H 
t^ 
•H 

Q 



-P 

P 

•p 

p 
o 

tr> 


H 

rd 
C 

< 



Eh 

O 
-P 

o 

<D 

-p 


a 



Volume 

Time Mean Speed 
Mean Time Headway 
Mean Occupancy 


Volume 

Time Mean Speed 
Mean Time Headway 
Mean Occupancy 


i 

Volume 

Time Mean Speed 


— 

Occupied-not 
occupied Status 


Occupied-not 
occupied status 
for each loop 




j 


Time and Duration 
of actuation 


Time and Duration 
of actuation for 
each loop 


Speed and Time 
of Actuation 


r ■ ~ ■ - 1 

i 

j Single Loop 


j Coupled Loops 


Doppler Radar 



26 



freeway traffic systems. To service these peripheral func- 
tions and to provide the user with the output information 
best suited to describe each subject traffic situation, a 
variety of options are made available: 

• The standard output (consisting of such MOE 
as vehicle-miles, vehicle-minutes, volume, 
density, speed, delay per-vehicle, lane-changes, 
etc.) will normally be reported at the end of 
each simulation subinterval, on both a link- 
specific and network-wide basis. The user may 
also elect to generate these reports at 
specified time intervals within each subinterval. 
These statistics are cumulative either from the 
start of simulation or, optionally, from the 
beginning of each subinterval. 

• The output of surveillance detector data may be 
restricted to individual links or inhibited 
altogether. This option permits more economical 
use of the simulation model without major 
revision to a working data deck. Detector data 
output by INTRAS may either be analyzed immediately 
or retained and analyzed at a later time. This 
procedure permits re-analysis with modified point 
processing or system analysis algorithms without 
costly repetition of the simulation process. 

• Data may be output to tape or disc file for 

later processing by a plotting module. Through this 
option, vehicle trajectory and/or MOE contour 
plots may be created for selected freeway links 
or groups of links. 

• The MOE data in the standard output reports may 
be saved for processing by the statistical 
analysis procedures. It will be possible to 
compare pairs of simulation runs, performed to 
study the effect of parametric variations, for 
statistical agreement by use of these procedures. 

• It is possible to specify one station,, on each 
freeway or ramp link, at which lane-specific mean 
values and distributions of headway and speed will 
be reported. These statistics, plus lane-specific 



27 



volume, will be output in addition to the standard 
report at the end of each simulation subinterval. 

• Values of embedded calibration parameters (i.e. 
grade dependent speed or acceleration, lane 
distribution of desired free-flow speed, etc.) 

may be varied via card input. Thus, as an example, 
it is possible to revise all characteristics of a 
particular vehicle type without recompiling any 
portion of the computer program. 

• The user may choose to execute only the data 
diagnostic procedures in order to check a new 
or revised data deck. This option provides for 
rapid data "check out." The more time consuming 
simulation may then be run under a lower priority 
to realize cost savings. 

The above user options ensure that INTRAS is capable of 
efficiently satisfying the requirements of a wide variety 
of research studies. 

3. THE STRUCTURE AND METHODOLOGY OF INTRAS 

The INTRAS simulation model is designed so as to satisfy 
a set of criteria which reflect the project goals and prior 
experience in designing and using traffic simulation models. 
These criteria may be stated as follows: 

• The INTRAS program design must include provision 
for all functional capabilities specified in the 
project work statement and described in Section 2. 

• INTRAS must be capable of realistically and 
accurately representing the dynamic and stochastic 
nature of traffic in a freeway-ramp-service road 
system. 



• 



The program must be designed for maximum utility. 
The user must be able to input data and exercise 
program control options via clear, concise input 
and program execution procedures. 

Programming and internal data must be designed so 



28 



as to minimize the total core storage required. 

• To the extent possible, after consideration of the 
preceding criteria, execution time must be minimized 

The following sections describe the structure and 
methodology incorporated into the INTRAS design to 
achieve the above objectives. 

3. 1 Functional Structure of INTRAS 

The INTRAS model is a highly complex system con- 
taining procedures for multi-purpose input processing, di- 
agnostic testing, microscopic traffic simulation, output 
reporting, statistical analysis, detector output processing 
and digital plotting. Reliable and efficient use of such a 
system depends on the system structure and organization. 
Early in the development of INTRAS, the time ordering of 
the above procedures was determined. The independent por- 
tions of each procedure were identified and isolated. 
This planning function provided a basis for the structural 
design of INTRAS. 

One danger encountered in the implementation of large 
complex programming systems is that, when they are com- 
pleted, storage restrictions may limit their usage to 
trivial applications. To achieve meaningful results it may 
then be necessary to segment the system. This unplanned 
segmentation often leads to loss of efficiency and debili- 
tates certain features of the original plan. It is of 
utmost importance to design a system functionally, so that 
segmentation is planned and the logical flow between 
segments occurs in the most efficient way, consistent 
with project goals. 

INTRAS is designed with these factors in mind. The 
functional segments are overlayed so as to provide maximum 
computer storage for arrays and scalars. The maximization 
of data storage assures the greatest utility from the 
standpoint of treating large traffic networks. Major por- 
tions of the data storage are optimized to increase network 
size capabilities to the maximum. This optimization is 
discussed in Section 3.3. 

The overlay structure and information transfer of INTRAS 

29 



is illustrated in Figure 6. The functions associated with 
each module are described in detail in the following sub- 
sections. 

3.1.1 The INTRAS Supervisor Module 

The INTRAS system will exercise its various 
functions by performing transfers (CALLs) to the functional 
modules in the order dictated by input Run Control specifi- 
cations. The logical entity performing these transfers is 
referred to a s the INTRAS Supervisor. A flow chart of the 
Supervisor logic is included as Figure 46 in Appendix A. 
To ascertain the first module transfer of each case, para- 
meters on the Run Control card (for that case) are inter- 
preted. Thereafter, before returning to the Supervisor, 
each module determines the next necessary transfer and 
communicates this choice via a control variable (NEXCAL) . 

There are a number of service subroutines in the Super- 
visor overlay module which are referenced from many loca- 
tions in several of the overlay modules. Since they must 
reside in storage at all times, these routines must be 
included in the Supervisor. 

Identification of the INTRAS Supervisor module sub- 
routines, and a brief description of each, is presented in 
Table 2. In addition to the routines identified in Table 2, 
the following service routines which perform packing and 
unpacking of individual data array elements are included 
in the INTRAS Supervisor: 

Array Unpacking Routine Packing Routine 



LNKR 


ULNKR 


PLNKR 


LNKS 


ULNKS 


PLNKS 


VR 


UVR 


PVR 


VS 


UVS 


PVS 



3.1.2 The PORGIS Module 

The PORGIS (Program to ORGa nize the Input 
Stream) module is a pre-processor which reads all simula- 
tion oriented input data and performs an exhaustive series 
of diagnostic tests thereon. Errors uncovered by these 
testing procedures are reported and tables of input data 
are output. 



30 



r \ 




Ira n I 


W 


U ; 


S 


f D 11 / 
4J -I-l 4J / 


Eh 


c « o I 


!3 


O 1JH 


I— i 


u -u a \ 






a) 




ct> 




id 




M 









4J 




W 




C 







1/1 


■H 


0) 


4-> 





01 


•H 


e 


> 


u 


(!) 





D 


M-l 




C 


h 


HI 






o 

Sh 
0) 
4-1 

w 

CJ 

fd 
j-i 

En 

C 

o 

-H 
+J 

(d 

S 

Sh 
O 

M 

C 
(0 

0) 
Sh 
P 
+J 

u 

D 
U 
-P 



rd 
C 
O 

-H 
+J 

u 

c 

fa 



CD 

tn 

•H 

fa 



31 



Table 2 



INTRAS Routines 



Routine 



Function 



INTRAS 



Main supervisor program. Reads Type 99 
card and performs routing for each case by 
calling the appropriate overlay modules. 



EROT 



Generalized error message generator. This 
routine reports each error condition by 
number and prints the associated parameters 
(See Section 3.6). 



DRWS 



Disc read/write subroutine. This subroutine 
performs most of the I/O processing re- 
quired by the other subroutines. Error 
messages and standard reports are not 
generated by DRWS. 



UNPAK 



This subroutine unpacks one entire parameter 
vector for any of the arrays LNKR, 
LNKS, VR, VS (see Section 3.3). 



PAK 



This subroutine packs one entire parameter 
vector for any of the arrays LNKR, 
LNKS, VR, VS (See Section 3.3) . 



FINDL 



This subroutine locates the link number 
and type of a link described by its upstream 
and downstream nodes. 



BLOCK DATA 



Defines variables and arrays related to 
the INTRAS model structure. 



FINDLN 



Determines lane alignment of two freeway 
links or a ramp and a freeway link. 



NEXTLN 



Determines the downstream link, link type 
and lane for a given freeway or on-ramp 
link. 



LRANK 



Order freeway links, downstream to up- 
stream, so that downstream links may be 
processed first. 



32 



PORGIS also performs some data service functions. Sim- 
ulation case data which is error free may be stored on a 
Case Data Tape for future execution. A table of contents 
for this tape, displaying identification for each case, may 
be generated as a user option. 

Identification of the PORGIS module subroutines, and a 
brief description of each, is presented in Table 3. The 
logical flow of PORGIS is illustrated as Figure 47 in 
Appendix A. 

3.1.3 The LIS Moduel 

The function of the LIS (Load Input Stream) 
Module is to read and load simulation case data into the 
model data arrays and scalars , to perpare for execution of 
each simulation subinterval. The diagnostic tests of 
PORGIS are not performed, as LIS is not referenced unless 
the data deck has been shown to be error free. Such is the 
case for case data, previously checked by PORGIS, in a 
previous computer run, and stored on the "Case Data Tape" 
(see Figure 6) . The logical flow of LIS is illustrated as 
Figure 48 in Appendix A. All subroutines of the LIS module 
are directly analogous to those of PORGIS with the diagnos- 
tic tests removed. Table 4 presents the correspondence 
between the subroutine names of LIS and of PORGIS. The 
simulation module, SIFT, is always specified as the next 
overlay to be called by the Supervisor, upon completion of 
processing by LIS. 

3.1.4 The SIFT Module 

SIFT (SImualtion of Freeway Traffic) is the 
overlay module of the INTRAS system which performs all simu- 
lation activities. Reflecting the scope and complexity of 
the simulation model, it was necessary to segment this 
module into two lower level overlays in order to satisfy 
the specified storage constraints. To avoid excessive com- 
puter time expenditures, for "spooling" these segments in- 
and-out of core storage, those routines which perform fre- 
quent processing activities (at least once per time step) 
are grouped into one segment. The other segment is com- 
prised of those routines which perform relatively infre- 
quent processing activities (i.e., output, fill time equil- 
ibrium testing) . The SIFT module subroutines are described 
briefly in Table 5. Logical flow diagrams for the SIFT 

33 



Table 3: PORGIS Routines 



Routine 



Function 



PORGIS 
Supervisor 



This main program of the overlay reads 
and checks the type 00, 01, 50, 51, 56 
57 and 60 cards. Subroutines are called 
to read the other card types and print 
input tables 



T ABC ON 



LPAK 



Prints a "table of contents" for the "Case 
Data Tape" . 

Performs the function of compressing or 
expanding the LNKF, LNKR or LNKS arrays to 
either economize on storage or provide for 
a larger network (see Section 3.3). 






LINKIN 



INT1 



Reads and diagnostically checks the link 
geometry, name and operation cards (Types 
02, 03, 04, 05 and 06). The appropriate 
arrays and scalars are primed. LPAK is 
called to compact the storage arrays. 

Performs system wide geometric checking 
and primes certain system geometric 
dependent variables. 



TURN IN 



PRSIG 



PRMSND 



Reads and diagnostically checks the turning 
movement cards (Type 08) . 

Reads and diagnostically tests the signal 
and sign control cards (Type 10) . 

Reads and diagnostically tests the input 
demand volumes (Card types 20 and 21) . 



34 



Table 3: PORGIS Routines (continued' 



SURVIN 



Reads and diagnostically tests the sur- 
veillance detector specifications on 
card type 25. The DTCTR array is primed, 



INCIN 



Reads and diagnostically tests the in- 
cident specifications on card type 30. 
The INCID array is primed. 



IMBED 



Reads and diagnostically tests revisions 
to imbedded calibration data (Card Types 
35-49) . 



PRACT 



Determines if intersection actuated 
traffic control input cards (types 15, 16 
and 17) are present and calls appropriate 
subroutines to process them. 



CTPFF 



Reads and diagnostically checks the 
actuated controller cards (type 15) 



CTPSX 



Reads and diagnostically checks the 
phase cards (type 16) . 



CTPSV 



Reads and diagnostically checks the 
phase operations cards (type 17) . 



CLRALL 



Initializes COMMON variables and arrays 
before processing each case. 



INACT 



Prints a table of intersection actuated 
traffic control parameters for each traf- 
fic actuated intersection. 



DETGEN 



Generates synthetic detector stations 
required to output plotting parameters 



LINOUT 



Prepares output tables describing the 
link specific geometries and operation 
parameters. 



35 



Table 3: PORGIS Routines (continued) 



SI GOUT 



Prints a tabulation of the signal and sign 
control at each network node. 



FLOOUT 



Prints a table of demand volumes at each 
entry link. 



SUROUT 

INCOUT 
IMBEDO 



Prints a table of user defined surveillance 
detector specifications. 

Prints a table of incident specifications. 

Prints a table containing the current 
values of the imbedded calibration para- 
meters if any values have been revised via 
input. 



MATCH 



Tests for equality in the last two digits 
of 700 and 800 series node numbers. 



CHKNOD 



Performs validity tests on node numbers. 



36 



Table 4 : PORGIS and LIS Subroutine Correspondence 



PORGIS 
Subroutine 

PORGIS 

LPAK 

LINKIN 

INT1 

TURN IN 

PRSIG 

PRMSND 

SURVIN 

INCIN 

IMBED 

DETGEN 

LINOUT 

SIGOUT 

FLOOUT 

SUROUT 

INCOUT 

IMBEDO 

CLRALL 

PRACT 



Corresponding LIS 
Subroutine 

LIS 

LLPAK 

LLINKI 

LINT1 

LTURNI 

LPRSIG 

LPRMSN 

LSURVI 

LINCIN 

LIMBED 

LDETGE 

LLINOU 

LSIGOU 

LFLOOU 

LSUROU 

LINCOU 

LIMBDO 

LCLRAL 

LPRACT 



37 






Table 4: PORGIS and LIS Subroutine Correspondence 

(continued) 



PORGIS 
Subroutine 

CTPFF 

CTPSX 

CTPSV 

IN ACT 



Corresponding LIS 
Subroutine 

LCTPFF 

LCTPSX 

LCTPSV 

LINACT 



38 



Table 5: SIFT Routines 



Routine 



Function 



Supervisor 
or Low or 
High Overlay 



SIFT Executive which contains a simulation 
Supervisor control loop and calls low and high 
frequency overlays as required 



VPAK 



LASTLK 



Reallocates unused vehicle array storage 

Determines the previous link, link type 
and lane for a given freeway lane 



FRSTV 



Determines the most downstream vehicle 
in lane of link specified 



LASTV 



Determines the most upstream vehicle in 
lane of link specified 



FINDFV 



Locates an empty slot in freeway vehicle 
array 



FINDRV 



Locates an empty slot in ramp vehicle array 



FINDSV 



Locates an empty slot in surface vehicle 
array 



RANDOM 

HICON 

UPSIG 



Pseudo-random number generator 

Executive of high frequency overlay segment 

Revises fixed time signal timing and calls 
appropriate routines to link with traffic 
responsive control algorithms. Also flags 
end of queue and cycle failures when necessary 



SIGACT 



Performs intersection actuated traffic control 
processing 



PDAFZ 



Subroutine to poll all detectors for inter- 
section actuated traffic control processing 
during an active phase 



39 



Routine 



Table 5: SIFT Routines (continued) 



Function 



Supervisor 
or Low or 
High Overlay 



PDNAFZ 



Subroutine to poll all detectors for 
intersection actuated traffic control 
processing during an inactive phase 



UPACT 



Controls updating of signal settings for 
intersection actuated traffic control 
processing 



GRNSIG 



Updates green signal settings for inter- 
section actuated traffic control 



REDSIG 



Updates red signal settings for inter- 
section actuated traffic control 



H 



TERMFZ 



Terminates old phases and activates new 
phases during intersection actuated traffic 
control processing 



H 



DECACT 



Determines whether any phase now in green 
should enter clearance interval (amber) 
during intersection actuated traffic 
control processing 



H 



TERM 



Terminates phases during intersection 
actuated traffic control processing 



ACTFZ 



Activates phases, calculates interval 
durations and sets status codes during 
intersection actuated traffic control 
processing 



ASIG 



DETSW 



Sets signal codes facing entry links under 
intersection actuated traffic control 
processing 

Implements detector switching, if appropri- 
ate, during intersection actuated traffic 
control processing 



40 



Table 5: SIFT Routines (continued) 



Routine Function 

FZCL Determines if there is a call to a phase 
for a specified controller during inter- 
section actuated traffic control processing 

CALl Performs clock time ramp metering 

CAL2 Performs demand/capacity ramp metering 

CAL3 Performs speed control metering 

CAL4 Performs gap acceptance merge control 

CAL5 Performs clock time diversion 

CAL6 Performs least time path diversion 

CAL8 ) Two dummy subroutines included so that 
CAL9 / additional traffic responsive control algo- 
rithms may be integrated 

MOEV Loops over all lanes of surface and ramp links 
calling appropriate subroutines to process 
non-freeway vehicles 

MOOV Processes all vehicles in a designated lane 
of a non-freeway link for the current time 
step 

SVEH Generates vehicles on surface entry links 

GOQ Processes vehicles eligible to discharge 
from surface or ramp links 

OFFRMP Interface between freeway and off ramps 

HDWY Calculation of queue discharge headways 
(non- freeway) 



Supervisor 
or Low or 
High Overlay 



H 

H 
H 
H 
H 
H 



41 



Table 5: SIFT Routines (continued) 



Routine 



Function 



Supervisor 
or Low or 
High Overlay 



GETCD 



Generates turn codes for vehicles entering 
new links 



LSWCH 
LANES 



Perforins non-freeway lane switching 

Assigns discharging vehicles to lanes on new 
non- freeway links 



TSTSAT 



Tests for saturation on links receiving dis- 
charging non- freeway vehicles 



DETECT 



Revised detector array to provide current 
status each time step 



TSIG 



Examines signal code and decides whether it is 
permissive or prohibitive for a given vehicle 



NORM 



Computes normal trajectory for given vehicle 
for current time step 



H 



CLNUP 



Performs bookkeeping and statistical updates 
at end of each time step 



INCDAT 



Outputs data to "INCES Data Tape" at each 
vehicle actuation of a surveillance detector 



H 



TPTOUT 



Outputs data on vehicle trajectories to "INPLOT 
Data Tape" at user specified intervals 



QSTATE 

TYPE 

SELEN 

FILL 



Determines if a particular non-freeway vehicle 

is in queue at end of time step H 

Generates vehicle type or driver type H 

Determines environment of non-freeway link 

being processed H 

Fills non-freeway link environment arrays H 



42 



Table 5: SIFT Routines (continued) 



Routine Function 

RELEN Updates link array parameters for receiving 
and approach links in environment of sub- 
ject link 

SEVEN Creates environment of vehicle being 
processed 

GETUNV Determines which vehicle unpacking routine 
to call by testing type of link being pro- 
cessed 



Supervisor 
or Low or 
High Overlay 



H 



REVEN 

ONRMP 
FMAIN 



Updates vehicle array parameters, after sub- 
ject vehicle has been processed, for vehicles 
in environment of subject vehicle 

Interface between on-ramps and freeway links 

Loops over freeway links to process vehicles 
for freeway time step 



H 
H 



CLOSE 



Determines points at which a lane will 
be closed due to incidents or lane- 
end 



BLOK 



Determines whether a vehicle can stop before 
reaching nearest blockage and adjusts new 
speed if necessary 



CONSOL 



Consolidates a sequence of CALLS to other 
lower level freeway processing routines so 
that they may be referenced from multiple 
locations 



FMOVE 



Moves all vehicles in one lane of a specified 
freeway link 



EMGNCY 



Calculates accelerations for collision 
avoidance of freeway vehicles 



43 



Table 5: SIFT Routines (continued) 



Routine Function 

FGNRAT Generates vehicles on freeway entry links 

CHANGE Determines target lane for freeway vehicles 
desiring to change lanes 



Supervisor 
or Low or 
High Overlay 

H 



CHECK 



Determines if change to freeway target lane 
is currently possible 



H 



CHOOZ 



FRESET 



Determines exit lane, or desire of lane change 
or yielding-way on freeway 

Updates the identification of the last vehicle 
in the specified lane of the specified freeway 
link 



H 



H 



RISK 



Determines the maximum deceleration acceptable 
to a freeway vehicle changing lanes or of its 
new follower 



H 



LCROSS 



Updates parameters after freeway vehicle 
crosses link boundary 



ALANE 

CANCEL 

COLECT 

ADVANC 

LOCON 

INIT 

RESET 



Identifies the origin lane of a freeway lane- 
change maneuver H 

Modifies the tracing (leader- follower) linkage 

for freeway vehicles H 

Updates detector and freeway data station arrays 

to reflect current status H 

Implements "Advanced Warning" sign logic H 

Executive of low frequency overlay segment L 

Initialization of parameters outside time step 
control loop L 

Resets statistics at end of network priming 

period (fill time) L 



44 



Table 5: SIFT Routines (continued] 



Routine 



xr.- A. 



Supervisor 
or Low or 
High Overlay 



FILTST 



FTSC 



CPTOUT 



Determines whether equilibrium has been 
attained during fill time 

Calculates appropriate current freeway 
time step 

Outputs MOE data for contour plotting to 
"INPLOT Data Tape" 



CYCP 

INT ST 
SINCES 



Prints reports of cumulative or interval 
specific MOEs 

Prints intermediate (detailed) data reports 

Implements incident detection algorithms for 
feedback control applications 



L 

L 



SPOINT 



Performs point processing of individual 
detector data for feedback control applica- 
tions 



SINC1 
SINC2 
SINC3 



Performs incident detection algorithm 
processing for feedback control applica- 
tions 



45 



Suparvisor, HICON and LOCON are included as Figures 49, 50 
and 51 in Appendix A. 

The high frequency activity overlay is called from only 
one location in the SIFT Supervisor (at the beginning of 
each new time step) . This overlay is only reloaded if the 
low frequency activity overlay is in core. The low fre- 
quency overlay is called from several locations, depending 
upon the options requested; a flag, ISSEQ, indicates the 
logical path to be followed in LOCON. 

3.1.5 The POSPRO Module 

The POSPRO Module ( POS t PROc essor) is de- 
signed to act as an interface between the SIFT simulation 
and the statistical testing procedures of the SAM module. 
The task performed by POSPRO is to create a file of simula- 
tion statistics on the Statistical Data Tape. This module, 
consisting of a single routine, is entered at the end of 
each simulation subinterval to add the subinterval link- 
specific and network statistics to a temporary file. When 
the total simulation run is completed, the contents of this 
file are added to the Statistical Data Tape. 

A flow chart of the POSPRO logic is included as Figure 
52 in Appendix A. 

3.1.6 The SAM Module 

The SAM (Statistical Analysis Module) 
module is designed to perform statistical comparisons be- 
tween pairs of simulation runs. The purpose of such com- 
parisons might be to determine if a particular parametric 
variation results in a significant change in key MOEs . 
Since it is likely that the user may use INTRAS to resolve 
just such questions, SAM is included as part of INTRAS to 
satisfy this need in the most efficient manner. 

In addition to comparing the results of a pair of simu- 
lation runs, SAM may be used to compare simulation results 
with externally supplied MOE (i.e. field data). Data formats 
are provided for input of such external data to create the 
equivalent of a simulation run statistical file on the Sta- 
tistical Data Tape. A logical flow chart of SAM is included 
as Figure 55 in Appendix A. Table 6 presents a listing of 
the SAM module subroutines and a brief description of each. 



46 



Table 6 



SAM Routines 



Routine 



Function 



SAM 
Supervisor 



READCL 

PRINT 

STAT 

TTEST 

CALC 
ANOVA 



This main program of the SAM module reads 
the SAM data cards, performs file crea- 
tion or management activities, and calls 
subroutines to accomplish a statistical 
comparison between files. 

Reads statistical data from statistical 
Data Tape and primes arrays . 

Prints contents of arrays for cases to be 
compared . 

Performs statistical analyses by calling 
individual test algorithm subroutines. 

Calculates paired T-test statistics and 
determines significance. 

Calculates arithmetic means and variances 

Performs one-way analysis of variance 
tests . 



WILCOX 

UTEST 
RANK 



Performs the Wilcoxon matched pairs 
signed-ranks test. 

Performs the Mann-Whitney U-test. 

Algebraically ranks vector elements. 



47 



3.1.7 The INCES Module 

The purpose of the INCES ( INC ident Detec- 
tion and Estimation) Module is to perform all processing of 
detector data output. In addition to implementation of in- 
cident detection and parameter estimation procedures, this 
module includes point processing and MOE evaluation algo- 
rithms. INCES may be employed as an integral part of a 
simulation run or as a separate run to process previously 
stored data. In either case, detector data is taken from 
the INCES Data Tape, as primed by SIFT. 

A flow chart of the INCES logic is included as Figure 
54. Brief descriptions of the INCES subroutines appear in 
Table 7. 

3.1.8 The INPLOT Module 

The INPLOT Module prepares vehicle trajec- 
tory and MOE contour plots based upon data, stored by SIFT, 
on the INPLOT Data Tape. Both plot types are generated in 
the time-space plane. The space axis may represent one or 
a group of contiguous freeway links. The user may request 
vehicle trajectory plots, in a single lane, or for all 
lanes, in a specified section of freeway. The available 
MOE's considered in the contour plots are: spot speed, 
volume, density, delay/vehicle-mile, headway, and travel 
time/vehicle-mile, as specified by the user. 

Contours are plotted, for each MOE, at a standard set 
of values embedded in INPLOT arrays. These standard values 
may be revised via the INPLOT parameter card. In addition 
to this option, an index may be created which details the 
current contents of the INPLOT Data Tape. 

A flow chart illustrating the INPLOT logic is included 
as Figure 53 in Appendix A. The identification and a brief 
description of the INPLOT subroutines appear in Table 8. 

3.1.9 The FUEL Module 

Recently, a new overlay module was designed, 
for the UTCS-1 model, which calculates fuel consumption and 
vehicle emissions data based upon the dynamics of indivi- 

48 



Table 7 : INCES Routines 



Routine 



Function 



INCES 



This main program of the INCES Module reads 
the INCES Parameter cards and selects the 
proper subroutines to call to execute the 
prescribed procedures. 



READET 



Reads the detector data from the INCES 
Data Tape 



POINT 



Performs detector specific (point processing) 
parameter evaluations 



POUT 



Prints a report containing the results of the 
point processing evaluations. 



INC1 
INC2 
INC3 



Performs the incident detection algorithm 
processing 



PINC 



Prints a report containing the results of 
the incident detection algorithms 



M0E1 
M0E2 
M0E3 



Perform the measure of effectiveness evalua- 
tion processing 



MOUT 



TTIME 



Prints a report containing the results of the 
MOE evaluation algorithms 

Implements the travel time algorithm for a 
number of MOE alaorithms 



49 



Table 8: INPLOT Routines 



Routine 



Function 



INPLOT This main program of the INPLOT module reads the 

Supervisor INPLOT request card and calls the appropriate sub- 
routines to perform the requested activity. 

AXPLOT Determines the extent of and plots the axes required 

for each plot. 

COMPCT Manages the trajectory data buffer. As the data is 

plotted COMPCT retains enough of the previous buffer 
to ensure continuity with the new buffers data. 

CONTR Reads MOE data, computes MOE values to be plotted 

and then calls CPLOT to plot. 

CPLOT Determines scaled time and space coordinates, at which 

MOE contour is to be plotted and then calls the 
CALCOMP software routines to perform plotting operation. 

IOPROC Reads data from INPLOT Data Tape and splits it into 

two temporary files for contour MOE and vehicle 
trajectories. 

PATH Identifies and sequences those links and lanes from 

which data is required for a particular plot. 

SEARCH Searches trajectory data to find matching vehicles 

for successive time steps. 

SPTAL Identifies each contiguous lane segment in a freeway 

section and calls TRAJEC to plot trajectories for 
each segment. 

TPLOT Scales trajectory time and space coordinates and 

calls CALCOMP software to perform plots. 

TRAJEC Reads trajectory data and calls appropriate sub- 

routines to plot vehicle trajectories. 

PAGE Draws page outline, writes plot heading and roadway 

section indicators. 

FILEX Prints tape file index. 

LEGEND Draws a legend of contour plot symbols. 



50 



dual vehicles at each time step. This overlay module, FUEL* 
is included in the INTRAS design, as a user option, to pro- 
vide link-specific evaluations throughout the course of a 
simulation run. Calculations of fuel consumption and emis- 
sions (for carbon monoxide, hydrocarbons, and oxides of 
nitrogen) are based upon vehicle type and trajectory data: 
acceleration/deceleration and speed. These data are stored 
by SIFT on the Vehicle Trajectory Tape, each time step, 
for all network vehicles. MOE ' s values are reported at the 
end of each simulation subinterval by the FUEL Module. Fuel 
consumption and emission rates are specified internally as 
default data table which may be over-ridden by the user. 
The default data tables are representative of the following 
assumed characteristics for the INTRAS vehicle types: 

High Performance Passenger Car - 8 cylinder 
Low Performance Passenger Car - 4 and 6 cylinder com- 
posite 
Intercity Bus - Diesel powered 
Single Unit Trucks - Gasoline powered 
Trailer Truck Combinations - Diesel powered. 

Tables defining the response surfaces of emission rates 
and of fuel consumption rates, in the speed-acceleration 
plane, for each vehicle type, will be supplied by FHWA. 
These tables are a part of the FUEL module, but may be 
over-riddern by user supplied card input, as for the modi- 
fied UTCS-1 program. This module consists of a single main 
program plus a series of BLOCK DATA routines defining the 
tables . 

3. 2 The User Interface 

The most obvious, and therefore most sensitive, 
characteristics of any program are contained in the user 
interface. The utility of a computer program is often 
judged on ease of use, and quality and clarity of output. 
Because of the comprehensive nature of the INTRAS simulation 
and the wide range of run control and output options avail- 
able, it is particularly important that the user is provid- 
ed with orderly and unambiguous utilization procedures. 

3.2.1 Data Input and Run Control Procedures 

The INTRAS input is patterned after its 

51 



forerunner, the UTCS-1 model (Ref. 4). To organize the 
input design process, functional data categories are de- 
fined. Within each category data elements are identified 
and then further subdivided so as to group those elements 
which pertain to particular subject areas (i.e. signal 
control, geometry, output specification, etc.). Table 9 
presents the hierarchy of data categories. 

A generalized card format is adopted to avoid prolifer- 
ation of FORMAT statements throughout the input routines. 
For the standard 80-column data card, this format consists 
of twenty-six three-column data fields followed by a two- 
column field for card type identification. All alphanu- 
meric data is grouped on the first few card types to permit 
unrestricted use of the generalized input format for the 
majority of the data deck. 

Data elements are assigned to card types by data cate- 
gory. Card type numbers (other than card type 99) are 
assigned to conform to the required order, in the input 
stream, of each data category. Table 10 identifies all 
frequency (i.e. number of cards for each application of the 
program) for each card type is also indicated. 

3.2.2 INTRAS Model Output 

The INTRAS Model produces many standard 
and optional output formats. The following subsections 
identify, describe and illustrate each major variety. 

3.2.2.1 Input Parameter Reports 

Tables of input parameters are 
provided, by the PORGIS and LIS modules, for each simula- 
tion case run. These tables fully identify the geometric, 
control and traffic input descriptors which characterize 
the current study. The values of parameters which may 
change with time are output each subinterval. 

Tables 11, 13, 15, 17, 19 and 21 illustrate the Input 
Parameter Reports for the traffic network of Figure 1 and 
2. Nodes 1, 2, 4 through 8, 13 and 14 have been excluded 
from Table 15 for brevity. The format of their sign con- 
trol output would be similar to that of node 12. Tables 
12, 14, 16, 18, 20 and 22 contain definitions of the column 

52 



Table 9: INTRAS Data Categories 



Data Category 


Subcategories 


Run Control 


Run Oriented Control 




Simulation Oriented Control 




Subinterval Oriented Control 


Network Descriptors 


Geometry 




Traffic Control 




Surveillance 


Traffic Descriptors 


Volume 




Routing 




Incidents 


, Output Control 


Printed Reports 




Detector Output 




Raw Data for Plotting 




Storage of Statistics for 




Comparisons 


Calibration Revisions 


Imbedded Parameters and 




Arrays FUEL Data Tables 


Parameters of Peripheral 


Detector Data Processing 


Function Algorithms 


Statistical Analysis 




Plot Generation 



53 



headings on the corresponding Input Parameter Reports. 

3.2.2.2 Standard SIFT Output 

The standard output report formats 
of the SIFT Module are illustrated in Tables 23, 25 and 27. 
These reports may be either cumulative or subinterval spe- 
cific. The latter form is shown. The samples represent 
the statistical results from one five-minute simulation sub- 
interval for the network of Figures 1 and 2. Definitions 
of the column headings for these three output reports are 
presented in Tables 24, 26 and 28. 

3.2.2.3 INPLOT Module Output 

The INPLOT Module creates CALCOMP 
digital plots of vehicle time space trajectories, and con- 
tour maps of MOE values in the time-space plane, as re- 
quested. Figure 7 is an idealization of the vehicle tra- 
jectory plot design. The actual INPLOT output would con- 
tain the trajectories of all vehicles traversing the de^ 
signated roadway section. To assist the user in correlat- 
ing the plot with the roadway geometry, the position of 
nodes within the designated roadway section is displayed 
along the horizontal (distance) axis of the plot. 

Figure 8 illustrates the contour map output capability 
of INPLOT. This output form may be produced for several 
different Measures of Effectiveness including: spot speed, 
volume, density, delay, headway and travel time. Contours 
are produced which represent constant values of these MOE's. 
The default family of speed contour values are displayed in 
Figure 8. A unique symbol is associated with each value 
and is used to label the plotted contours. The MOE contour 
values may be updated, as a user option, by card input. It 
would, for example, be possible for the user to specify a 
family of speed values of 50, 51, 52, 53, 54, 55, 56, 57, 
58, 59, 60, 61 and 62 MPH in order to obtain a much finer 
contour map than that shown in Figure 8 . 

As for the trajectory plots, node numbers are displayed 
along the distance axis for reference. 

3.2.2.4 INCES Module Output 

The results of processing surveil- 
lance detector output is reported by the INCES module in 
three formats, illustrated in Tables 29, 30 and 31. Table 29 

54 



g 

0) 
M 
+J 
CO 

•P 

3 
Ch 

C 
H 

o 

C 

o 

•H 
4-> 

•H 
M 
O 

J) 

a> 
a 



o 

<U 
rH 
XI 

CO 

E-t 



























u 


Li 


1 


























3 


U 






Li 




kj 




U 




V| 













C. 


1 




01 









u 




"J 










O 
















a 




a. 













X 




> 


























C 




u 


>i 




>. 




>■ 




X 










>• 


•H 




c 


'M C 




U C 




u c 




c c 










10 




c 


0) 

















■rt 










e 







3 


■W -i 




- -H 




JJ -H 
















U 




CT 


10 w 




(3 -i 




13 4J 




4J 















JJ 


U 


•c -o 




■O 13 




•a "J 




u .-o 












B. 


<3 


li 


c « 




C -J 




C -i 




u — 










a 




fH 


fa 


10 - 

E E 

t> oi 

c 
o 




13 3 

g e 

11 01 

c 
o 




13 3 

5 S 

■H 

11 01 

8 




C 3 

S 

D -H 
01 












- 
a 

4J 

o 


91 


s 




3 
E 

■H 

01 

i 






a 






























c. 






























a 


>l 












>. 
















*J 01 


•3 












u 













j 






3 












13 




•-• 












u r 


1) 














Jt 


jj 












■H JJ 


11 














C 


13 










oi 


4J 


C - 



























3 


a >. 


3 Uj 












X 

















C .a 


W 












3 




<w 












31 >. 


c 








91 




10 


01 


•H 






i 






10 ^ *3 


01 - 




* 




M 






-H 


J 













e oi a 






u 


01 


11 




u 


n 


C 












3 01 


£ 01 







e 


£ 







J 


11 












C O 01 


C JJ 01 




3 












-3 










31 


O — 1) 


3 (U 




JJ 


*> 


3 




01 














01 


> u 


K TO 




O 




C 




ii 


e 












10 


•ooo 


c o 




c 


e 




^ 


u 




a 








e 





"J ~ - 


M -H U 









11 


e 


•O Ji 


4J 


E 













w a a 


■h >, 2, 




11 


■H 


•a 




Oi u 




13 








•H 


u 





j: m 




£ 


JJ 







U 


X 


h 






1 


JJ 


01 ci 


4J S 11 


.u —i U 




JJ 


a 






c a 


11 


1 








a. 


C £ 


01 O JJ. 


jj 






N 







01 












•H 


~> 




G-. C w 






•H 




u 


c 


o 








1 


u 


■H O 


IS T O 


C 11 




C 




3 


10 


3 Z 


■H 





O 






o 


JJ C 


JJ 11 4J 


— T3 a 






10 


11 


%i 


— 


m 




= 




1 


a 


a <3 


3 U 


>l"« 11 




>* 




U 


u 


-. -3 


U-l 


■a 


13 






a 





TJ O 01 


<JJ -J 




*j 


4J 


«J 


3 


•u 3 


13 


a 


e 






a 


u 


4J 11 


•H — . 01 


TJ 


■H 




01 


01 


1 H 


u 


a 










V 11 


W 11) H 


4J U 1 


SI 


JJ 


s 






U U 


-1 JJ 




j^ 








V J3 


O ■" 


3 13 11 


u 




•H 


3 


b 


U X 


c 


c 


c 




i 




U -J 


m uj 


51 E 


01 












u ii 


Oi D 


or. 






1 




— a 


Cl JJ 


■a u •-■ 




-a 


>W 


u 




4- 


C •-< 










1 


01 £ 


3 «B 


•H JJ 


u 


■H 







^ 


01 


— 13 


in 








i 


ii s 


AJ T3 11 


u 


•J 






•a 


S 


T3 11 


> 10 




u 








-o 


a £ 


jj o e 


•9 

01 E 


jj 


■a 


c 


3 


U = 


-< U 


IU 






i 




01 


■u 01 01 




c 




13 


U 


-. 10 


41 ci 





u 








0) 11 


oi m «n 


E 


D 3 


11 


3 








u 




u 








1) -4 




£ U - 


-h e 


E 





e 




01 


1) rt 




s 




1 






01 <M .3. 


i! H.S 


u 


5) 


J3 


■3 


>. 


ii j: 


01 U 13 













<*J >H 


11 -rf q 


- E 


jJ 




U 




•o c- 


u u 




e 








■H 4J 


cun : 


13 £ = 


j: 


13 




U 


3 


3 


C 01 -rt 


ii 


10 








u e 


•H C 01 3 


JJ 3 


'J -0 


ii 





ii 


Cl 


S 


IB X J 


13 










OJ I! 


<W O U Ih 


01 C 11 


> c 


01 


a 


01 


u 


13 U 


-< C U 









1 




a i3 


"J T3 


JS 


13 




a 




u 


U £ 


•h a 













W M 


Q M 


< rt 5- 


" 


< 


= 


3 


fa 


3S c- 


- > 


J 


< 




1 







— . 


0) 




T 


jj 


J 














01 


3 




fa 


D 


Z 


9 


c 


10 




-1 


-J 


JJ 


a 





OS 


01 




u 






M 




JJ 


a 




c 


01 


s. 


c 


3 


10 


n 


3 


£ 


j 


frc 


X 



«« 41 > . 



C C 11 E 
•H -n u 3 

- - - z 



■! 


J 




.3 


■o 


•a 


11 




u 












10 


•i 


> 


01 


~: 





=: en 

C i-i 

_ - 



O M 

Si J 



22 



v ■> 1 












- - 1 




o 




i (N 




1 >1 


j* 


Q 


° 


! ° 


O 


y & \ 






1 


1 





55 



... 








^ 












3 


1 










- u 


3 




a 




J< 


■y s 






c 


£ 


C £ 










• u 


■H 


1 81 


01 






j<: <tj 


rt 10 








S 'J 


5) 


a 3 bj 


h 











91 




5 c 


iH 73 
91 


u t: 
a ai 


u -a ■§ 

9 2 .5 


- 


>! 


91 o 


au h 


<M 4J *H 


91 





91 ■•* 


g 13 «! 


u tj id 




73 


c 


U JJ 


«j o > 


3 91 > 





91 


u-i a 


u a n 


vi a u 


:!u 


c 






gj 91 


u u 




c 


U 3 


- U 4J 


M U U 


>i c 

3 S3 


'•J c 


<u 


O S 


a c 


o> . c 


41 


u 


B.-H 


a o -h 


a, 41 — 


a-- 




;■) 


j ^ 


J3 -2 






HI 


01 3 


11 3 


91 u 




>J 


U >. 0] 


<J >- 01 


-H 




C 11 


s <S 


C 13 


c 




o a 


o z 


C E 


c 


° 



I * 



3 
C 
•H 
■P 
C 
O 
U 



Eh 



r-l 01 01 

3 JJ 3 01 

«« 51 D 

— C 73 C 

i ,3 

4) -< — -H 



1, 4J — C O 
1) <u c 91 — 1 

2, 9) a J 
3 ■-( -H c IB 



e 

01 H 

Cl 4J 



5 I 



73 o> ■-< 



-1 £ C 

•J J 

~ 3 73 



» — kl 

3 -13 

I CP I 

u a 

r-l H -H 



■u O 

91 *J 

— Cr. 

C n 
O 

■H O 

JJ a 

•H 

V) 01 

o a 

■— 3 

nt 

T3 -I 



I _ 



CP AJ U 
rH U 01 

13 o 



>-l 01 <3 



P 






+J N 


u 






u ■- 


13 01 


c 




a -1 


-c K 


■H 




O 91 


'J 91 


01 




3 s 


01 u 


a 


01 


z. e 


— 91 




u 




■0 s 




a 




13 


a 


c 


c 



73 tf i 



H 


£ 


01 


O 




a 







01 






Uj 


O 


O 


-H 


iw 






I -2 J 



r;. 


IB 




01 




01 










91 






91 




u 















91 






O 


S-J 


U 


jj 








91 




O 




c 


-J 


V 


4J 


Uj 


r. 


91 


"J 


13 


ig 





E 


K 


T3 



1 








1 


H « 




1 


; c 






13 73 




18 


1 






C U 




U 




M 


CI 


tji 3 


■ 1 


91 73 


I +j 


= c 




■H O 




a U 


1 « 


■H 


■i^ *J 


Ul 




C "1 


- 




c c 


■H 




u 


91 


jj 


U "J ) 


73 O 


, O 


>l 


a 


91 13 01 


3 £ 01 i 


C k4 


! s 


13 c 


. 73 


U U 73 


6-. CI 73 ; 


13 4J 


: a : 


s 


! u 


13 91 U 


> U : 


3 


Z ! 


91 -H 


2. <a 


H-i C 13 


a « 


C O 




9) -u 


= u 


1- O O 


c s ; 


r^ 




u 


" 




■*H 1 






Cs. 


x 


Ul 


J 


w 



I 73 91 



56 



T3 

Q) 

C 
■H 
4-> 

C 

o 

u 



Eh 





c 


Ll 


m 






c 


Li 


in 


u 












Ll 










3 










3 


■<-4 






U 


*j 




•i 




U 


4J 




3) 


u 


IS 












■3 


O 


a. 


n 




u 


>< 


4J 


n 




Lj 








-u 


H 


3 1 




- 


±J 


r- 


jj 


= 


e 




— 


•y 


= 


3 








j :j L l, u 

-• c o o <e 



•i C O O in 



c 




JZ 




3 


c 






o 




U 


ij 






IC 




t) 










a 










>1 












Ll 


JJ 


■o 




>i 


:i 




13 





a 


A 


-1 


c 




aj 


> 






01 


3 


(3 


u • 




u 




6 


1) 


11 ' 




a 


M 




,1. 








u 


w 


11 


e ! 


13 








u 




C 


a 




14 




a 





V 


01 


V 


o 


3 • 


-H 


c 


u 


a 


a 


-■' 


■U 






U J3H 
3 3 3 
3 « 5 






£ I 



O u-i 3 
73 ul oi 



! 5 



o 


3" 








oi 








ij 


en 


n 


c 




3 


-» 



■H 4J ! 



C ic 

3 C H 

IC »H 

6 O 

13 oi u 

<u .* -) 

l, c c 

AJ -H G 



2. £ ; 

O J I 

s* I 

3 3 

£ 
u a 
■U 3 
C 

O -I 

a j .i 
<3 c ! 

■5" 01 
u ly — i I 
13 Li jj | 

"3 O i 

COO 



a e j 

01 -n 13 
•3-13 



Li 13 C 
11 —I C 

^ U 



jr. l. "3 oi 
j a. = 13 

CU -3 .3 



3 -i-i 

71 ii 3 
C O 3 



'J 13 •!-! 

13 J 

L, 3 

a 



ci 



:i 6- 



T3 13 r4 r} 



1 


pH 


r-. 


c ! 


<g 






«. 


■Ll 


O 




a 


O I 


























O 








o 


•J ! 










-> 


O 


a 


•n 


fl | 










B 


o c 


c 




u 


C 






3 


i-i 


2. fl 




Li 


a 









L 


X 


>,|H 


■n 





al 




>t 






u 


•" 






ii < 




13 




a 


11 








U) : 




3 




.3 


c 


C 


•H 







■n 






O 








1 




11 


i 






11 


s 


c 






AJ 


1 




3 


- 


— 







1 


u 


Li 






i 




—1 


1) 











1 




c 






in 1 


Uri 




01 




u 


> ••* 


■o 


J 


a. ' 















3 




O l 


13 




c 


) 




>. > 


13 





o ' 










>t 


J2 ~ 


1) 


■3 


| 


01 


-u 


ILl 


■U 


- 


•a -a 




Lt 


u 1 








-1 


TT 


O D 


13 





° I 


3 


V 


c 




.0 


■^ — 1 






| 






Q 





O 


14-1 U-l 


























X) 




Tl 


4J 




•1 


u o 


.* 


c 




J* 




M 








■J o 




c 


C 1 




13 


i3 


c 




C 3. 


•H 


•M 


■J 1 


—i 




4J 


T 





-H 3 3 

■-> O 3 

O — -w 

Li _ -a 

J -H Li 

ceo 



o o 'j i t i a u 



a ai I _i 







Li 


Li 







3 







^ 






- 


; 




"3 


11 


(3 






O 


] > 








JJ 













' V 










tn 




•Jl 






0! 


c 












C 


1 "3 






•3 


• 'J 


13 






13 





1 11 










1 C 


u 


a 




' "O 


u 


L. 


2 




Li 


01 




01 




1 13 


^ 


> 


3 






3 




■Ll 


_i 


13 




13 


-u 


1 £ 


u 


u 


U 


1 u 




'J 




U 




x: 




£ 




1 1) 




3 












< 




1 c« 








3 




•J) 




" 






a 


■Jl 




01 




CO 




1 * 




CO 




! "^ 
















HJ 








M 










3 


u 




: 'J 




o 




U 




^ 




c 




i 


■n 




tn 




•n 


ci 


CO 


a 


tn 




CO 




03 




1 S 


<-i 








O 








o 














— 


~ 


J 


li 


- 


c 


- 




J 


- 


- 





u cl ; j^ 



57 



T3 
Q) 

C 
•H 
-P 

c 



u 



a) 



























- 




i 








1 




l 


1 






i 


<3 




i a 






i 


<0 


1 


•3 


1 13 




i 


« 


> 




w J -. 




t4 


J_ 




U fi 






u 


j 




U U 




3 3 3 






3 


3 


3 3 


3 


3 3 3 


a 


z 


3 


3 01 




o in S 




U 


0) 




in 


£ 


o in £ 





(11 


E 


JJ 









'J 






u 







o 






C 




<J « 







JJ 


in 


JJ 


in 


jj in 





JJ 


in 


o — c 




HI 






in 




0) 




in 




in 




J3 


>. 


>■ Vl "fcc 




>1 


u 


UJ 


>, u 


>jj 


>> u uj 


=- 


- 




>. 3 -J 





fl -1 




<0 


m 





to — t 





■3—0 


13 







13 S) JJ 


B 








— 




2 w 




X <jj 


£ 


<*j 




S <0 


01 
















-J 








JJ — 


3 


• c <z 






= 


nj 


• c 


13 


• C 10 




C 


<3 


. in 3 




-*■-,> 








> 


rt •-< 


> 


-J -1 j> 






> 


H M E 


01 


<3 U 




a 






13 


u tn 


to u in 


10 




U 


13 '-J "j 


kj 


e -j a s 




c 


01 


01 c 


C 01 


01 c 


C 01 01 c 


C 





01 c 


c *j m 




O JJ 










JJ 


u 


JJ 


O JJ 








JJ 







■ICC — 






E 


C i-t 


•H C 


C -H 


■-J c c - 




e 


C ~J 


.* a ■« 




jj — ' JJ 




JJ 





•H JJ 


JJ 


•H JJ 


jj -t JJ 


4J 





-1 JJ 


jj — o 














a. 




a 


a. 






On 




o 




o 






o 




° 


o 






° 




JJ 










, 




>j 












31 "*- 










13 




a 












c 










U 




c 








M 


! 












•H 




3 






£ 


-a 




to E JJ 










44 




3 










i 


a c 3 s 














C7 






JJ 







-i £ 91 










JJ 




01 








jj 




o -i ■-. « 














M 






u 






■rt jj x u 










01 




<J4 









c 




- o ig ai 










•o 










31 







o> j = a 














•a 












> a 





















13 


JJ 




u » oi « 










it 




■j* 








13 i 




JJ -j 3 o u 










c 










1 


O 




« jj o ■•* oi 




c 










3 






01 






U B--tC.fi 











-4 












Uj 




•H — t JJ j3 £ 




















<JJ 


■-. 




uj <o y <o 




JJ 






T) 










■H 


JJ 




JJ u 




a 










13 O 






u 


c 


c 


U C 9 I 











ia 










0) 


01 





r* C JJ 














144 JJ 






a 


T3 C 




UJ 0! 01 13 










o> 




13 






in 


■H 


JJ 


"3 0-10 














C 3 








■-I 


a 


C 01 11 .1 














C -J 






44 


e jj 


- 


10 01 "C » C 




e 






o 




■H fl 









o u 


w 


0) & a 




01 






(3 




— > 








— 01 





£ in » jj a 




-3 






a. 




H 01 






a 


jj jj 


a 


son 




•H 






31 











u 


IB 1) 


o 


JJ C 'J -j 














SjTJ 






01 


jj -o 




3 >, U — U4 14 

3 JJ *JJ 
H 3 J< 1 J - 
■3 "c u « 

13 JJ 01 9) — 

in i >w rt -o 3 




c 






C 


jj 
u 
JJ 




gi tal 
ing an 






jj 
a 

a 

IB 


Jl 

JJ 
b C 
01 
J- TS 
O -H 




: £ « 2 < o 


*j 








n 




-H r* 






a 


01 O 




i-i U kJ 


oi 


10 










TJ -4 








JJ S 




■H -3 CT 


oi 



















u 


01 -• 




-J C "C 33 -4 


3. 


a 






a 




- - 









■o 




3 .3 c - j: 


•Jl 


01 






01 











UJ 


jj 




— (3 » "O I3 




jj 






jj 












> 




■h ti £ a jj 


C 


a 






<n c 




31 






Jl 


<3 UJ 




M £ JJ jj 3) B. 


3 













JJ 






3 


3 




JJ — j3 3> 3. 


'J 


a 






01 -H 












o -a 




in -j o c in 


c 








£ JJ 












01 01 




•H -H 01 O 














C -i 






13 


u in 




a x j < 




61 






B 




< a 






> 


! 



^3 

Eh 



I -5 



01 - 

"j in 

SI — 3 
U -■ 

C -H 'J 

- j: £ 

3 01 

E- > 



! a £ 



01 JJ 
■i-. (3 -H 

it: jj 



I JJ 01 






E 








01 


JJ 


jj 


M 


.3 


-H 


01 


u 


M 


— 



1 








' 
























£ 


"O 








B 








J. 






















1 










u 








a 




I 


1 


£ 


1 




















M 










ig 
















01 




















I 


^ 


■" 


1 


1 




o 


1 


i 



















1 








c 











M 




o 


M 






o 















« 





















01 





io ; 





I 


C 


u 


o 


01 










j 


JJ 



















c 


JJ 





o 








01 


c 






03 


10 










c 


! fl 


~j 


JJ 


■0 




JJ 


















JJ 




3 






u 




i 


C "O 





' jj 


n 


01 


■j 


ID 


01 




a 


03 


c 




a 


< 


J 




Q 




M 






3 





J 




> " 


M 


O 


13 


M 


01 


01 






i 


01 








3 






O 


M 


j 


£ 




-3 


JJ 


- 


11 


3 


a 


01 




c 


JJ 


3 


•H 




-i 


3 


3 


B 


-J 








■3 


13 


01 T3 


10 




c 


cr 




c 


3* 




B 














14 




3 





o 


JJ 


u 


Z 


•H 





u. 


i ** 


JM 


01 


jj 


01 


01 




o 


■M 


10 







-w 


10 




O 




01 




3 




> J3 J3 


\ 


5 







o 


X 




T3 


u 


O j 


1 


T3 


J*J 




1 


■0 






3 


o ! 




11 


















c 








3 








C 






•J 


■3 






B 






1 * 






3. 






o 

















o 






- 






3 


« 






! 'J3 






■Si 






01 






j 


03 








O! 










1 


-J 


M 














































3 


01 






1 o 






vj 






a 








o 








U 












■3 


X 


03 




1 X 


■ji 




X 


03 






03 








71 






X 


VI 













c 






' o 


H 






M. 




3 
















o 












Z 




J 




! i 


J 




b 


J 




c 


- 






- 


J 






- 


- 










•3 01 








: 








































M S, 


Jl 





ci 


l ° 












in 








Ifl 








r^ 












13 > 


<n 


— 


^* 


1 " 






in 






L*l 








un 








in 












U 6> 


























1 




















i 



58 



i a, 


n 




! 1) 


tn 




1 ° 


9) 


c 


! r. 








o 









u 




1 • 


a 


id 


~* 







id 


U3 




1 c 


■tii 




l 


U 


a 



O O C i o o c 



-l ti- 



ed -I 

'J — 

s z. 

f a 






T3 
Q) 
=1 
C 
•H 
-M 

c 
o 
o 



I •- 



I s <« 



•d "0 

~i o 



-1 3 

xi o 



n 



'-. x. 
Q. O -< 
a 3* id 
-i *j xi 
C 4- — I -H 
■- -U >d 3* 



U I XI 

o ■ c '.l 

T> i !J 



xi 

■« o 

xi TJ 



-i U 3 



Eh 



o -) 

■-I 3 

xi C l 

3 xi xi — 

XI 3 J .13 

3 3 > 

T3 £ U 

xi xi 3 a 

-i o o -J -i 

.3 li : 

> >, >- -I -I 

U O (J '« J! 

D 3 3 3 

XI J 'J j£ « 3* 

C 3 3 U (0 

-i c r o xj -i 

Si V CJ 3 T. XI 

3 U ■- 3 



I -5 



U xi -I 
O <d 

SO ff 



f-H -U ^ 





a 







•d 




a 





u 


M 


XI 













-v 


d 




a 
















V 






3 


IX] 






< 
























XI 









•e 


v 


e 






C 





*1 


a 


u 


-i 


3 




1 







01 


IX 


Q 








XI 


•o 


X) 


UJ 


U) 


U 


u 


x» 


I* 



I iB T3 I .C 



! 2 5,, 



I Z 






c 


3 


t) 


r. 




o 


V 




a 








n 






xl 




o 


c 




'J 






- 


id 


W 


Q 


Ml 




X) 






C 




"I 


c 










U 


u 


a 




H 







Id 


■3 


x> 


i. 


X) 


4J 


u 



7 a i 








1 




1 •- - 




in 


^ 


r- 


S 33 


•a > | 




o 


o 




| >a ! 


|u 5. | 




1 


1 
1 


i 


1 1 



59 



a. | 






-< c. , — c 



! 5 



JJ o -j o 



T3 
<D 
P 
C 
•H 
-P 

c 
o 
u 



.-. rg 


TJ 


.2 u 


"J 


« 




JJ -i 




o u 


>< 


3 


01 (3 


— lu 


JJ —1 


13 ~ 


"J 


> 


— T> 


C 


B. 


u 






>, > 


JJ 


U 4J 


JJ fl 


— 


C 


JJ a 


~ 


u c 



TJ 








.1 


JJ -"J 




h 






c 









n 


c 




uj 





u 


U 3 






> 




u, 




51 











U 






>> 3 




D 


fl 




Ij 




-1 


HI 





-H 




t= 


a 


u 






3 






0] 


31 


C 


0) 


- 


5) rJ 








n 


•n 3 


■•J 





jj 


UJ 



0) U 



3". 5 



5 l 



Eh 




60 



T3 
<D 

iH 


c 
o 
o 





















10 


H 










> 


01 












a 










01 








01 


91 


01 -u 


01 








rt 


-t e 












•H -H 


•h 






w 


IH 


iw £ 


tw 




>1 






3 


*H 




u 


* 5 


> 


* in 


3 a 




c 


o> > 


V 


01 


01 > 





e i- 


c 


c l* 


C u i ; 




01 







» 


0* 


u u 


H 


u a. 


U JJ 


01 


V c 


01 


2U 


01 C ; 


u 


a-H 


g.i 


a--t i 


!u 


a 


c 


s 


a \ 




01 3 


01 -H 


0) H 


01 3 1 




y a 


U -I 


iH 


o m ! | 




c 


c 


c 


c 




o 


o 


° 





1 



-H | OJ 



! j- 






I e 



iH 
En 



2 i i 



I a 




| a 



3 ! 3 



s I 1 



— s ; ^ ia 



l* ' 
il 'J 



! = 


tfl 


! w 


1 M 


i '-" 


■ <fi 


!e • 






1 








m 




1 >" 


1 ^ 






r» 


CN 


1 » 


1 a% 








t 


1 


: 





61 



i 

< 
w 



u 

! 

< 



§ § § § 

[^ h fa fa 







I 



fa fa 



00 



o 
o 



o 

rH 

vo 



o 



O 
O 



00 

o 
r- 

\ 
o 
o 



o 



rr 

in 



o 
o 
o 

o 

o 

00 

o 

CA 



o 
o 

VD 



'3- 

in 



o 
o 
o 
o 

o 
o 

00 



00 



ro 


<N 


ro 


ro 


ro 


CM 


3 


CM 


ro 


«» 


in 




O 


r-t 


rg 


» 


■^ 


m 


r» 






ro 







00 



o 
in 



CM 



CM 



ro 



^ 



CO 



Table 12: Definition of Column Headings in 
"Freeway Link Definition" Report 



Column 
Heading 

L 

LINK 

LANE 

SPAN 

AUXILIARY LANES 

LGH 

A,D,F 



MEAN FREE 
FLOW SPEED 



Definition 

Freeway link index 

Upstream and downstream nodes of link 

Number of through lanes 

Link length (feet) , no value printed for entry links 



Auxiliary lane length (feet) 

The auxiliary lane numeric code (see section 2.1) will 
appear in the appropriate column to indicate acceleration 
deceleration, full auxiliary lane status 



Input value of desired free-flow speed (miles per hour 



GRADE 



PERCENT OF 



Imbedded calibration value of grade (percent) which 
most closely represents the input value 

Percent of traffic performing the movements left-turn (LEFT) 



VOLUME/DESTINATION no-turn (THRU) and riaht-turn (RIGHT) at the downstream 



NODE 



CURVATURE 



RAD 



P 




EL 




RT. 


LANE OF 


SEP 


. PAIR 


REC 


LANE 



intersection; followed by the node number at the downstream end 
of the next link. Turn percent and destination node number 
are separated by a slash, "/" • 

Indicates set of three parameters used to calculate 
limiting speed 

Radius of curvature (feet) . "0" indicates no value input, 
implying straight roadway 

Pavement condition code 

Superelevation (percent) 

Right lane of pair separated by physical barrier. Two 
such separations are permitted per link, hence, the FIRST 
and SECOND subheadings 

Lane in downstream "through" link receiving traffic from 
lane 1 of this link 



IDENTIFICATION 



DISTANCE FROM 
DOWNSTREAM NODE 



NODE LOCATING 
CFF-RAMP 



DISTANCE FROM 
OFF -RAMP 



Text describina link 

Position within link locating advanced warning sign (feet) 



Node at which off-ramp referenced bv advanced warnina 
sign begins 

Distance from advanced warning sign to off-ramp (feet) 



63 



<p H £ 



T 5 



¥ 

8 



T 



Dm 
I 



(A 

a 



C5 



si 



ES 



CM 


O 


o 


O 
O 
O 
rH 


o 


o 


O 


eg 

CM 


CM 
CM 



o t ^r 

(N <N 



in 


O 


o 


o 


o 


o 


o 


o 


o 


o 


o 


^r 


O 


o 


o 


o 


o 


o 


© 


o 


o 


o 


m 


O 


o 


c 


rH 


o 


© 


o 


o 


o 


o 


CM 


O 


o 


o 


o 


o 


o 


o 


o 


c 


© 


iH 


o 


o 


o 


*r 


o 


o 


o 


o 


o 


o 



CM 


CM 


CM 


O 
CM 


CM 


CM 


O 
CM 


o 

CM 


CM 


CM 


CM 

CM 


CM 
CM 


CM 
CM 


in 

H 


CM 

CM 


CM 
CM 


in 

rH 


m 


CM 
CM 


CM 

CM 



en 



w 



en 



0) 



o o o 



o o o 



o o o 
o o o 



o o o 



o o o 



o 
o 



o 



o 



o 
o 



en 

■H 
00 

s 

co 

co 



CO 



oo 
\ 

co 
co 



o^ 



o 

o 



o 
oo 

o 



£ 



o 
in 



oo 
in 



co 
oo 



o 
oo 



o 

CM 



o 
o 



o 
o 



0) 



0< 
en 



^ ^* ^* 

CM CM CM 



O O O 

o o in 

CO CO ^J* 



I 


CM 


CM 


o 
n 


CM 


CM 


CM 


o 

CO 


o 

CO 


o 
co 


o 

CO 


« 


o 


o 


o 


O 


o 


o 


o 


© 


o 


o 


J 


o 


o 


o 


O 


o 


o 


o 


o 


o 


o 


1 


o 


o 
in 
co 


o 
in 


o 
o 

CO 


o 
o 
co 


o 

O 
*3> 


o 
o 
co 


o 
o 


o 


o 



rH 

«» in 



r- 


00 


en 


o 

r-l 


rH 

rH 


CM 

rH 


O 


© 


co 

rH 




CM 

© 

00 


t> 


00 


en 


© 


rH 


co 

H 




VO 

O 
CO 


in 
o 

CO 



00 o\ 



64 



Table 14: 

Column 
Heading 

L 

LINK 
LANE 
SPAN 
POCK L R 



MEAN FREE 
FLOW SPEED 



GRADE 



PERCENT OF 

VOLUME/DESTINATION 

NODE 



TYPE OF DWNSTREAM 
INTRSECTN 



LOST TIME 



MEAN QUEUE 
DISCHGE HEADWAY 

CURVATURE 



RAD 

P 

EL 

ON/OFF RAMP 

REC LANE 

IDENTIFICATION 
OPP.LINK 



Definition of Column Headings in "Ramp 
and Surface Link Identification Report" 

Definition 

Ramp or Surface link index 

Upstream and downstream nodes of link , 

Number of lanes (excluding pockets for surface links) 

Link length (feet) , no value printed for entry links 

Capacity of Left and Right turn pockets (passenger car 
vehicle lengths) , for surface links only 

Input value of desired free-flow speed (miles per hour) 



Imbedded calibration value of grade (percent) which 
must closely represents the input value 

Percent of traffic performing the movements left-turn (LEFT) 
ho turn (THRU) and right-turn (RIGHT) at the downstream 
intersection; followed by the node number at the downstream 
end of the next link. Turn percent and destination node 
number are separated by a slash, "/"• 

Code indicating queue discharge characteristics at down- 
stream intersection 

Time required for first queued vehicle to react to green 
signal (tenths of a second) . A "0" indicates a distribution 
will be used, referenced by driver type. Not required for 
on-ramps 

Mean time between discharge of queued vehicles (tenths of 
a second) 

Indicates set of three' parameters used to calculate limiting 
speed 



Radius of curvature (feet) 
implying straight roadway. 



"0" indicates no value input, 



Pavement condition code 

Superelevation (percent) 

Indicates, for ramp links, which end connects to freeway 

Lane in downstream "through" link receiving traffic from 
lane 1 of this link 

Text describing link 

Link index identifying source of traffic opposing left- 
turners from this link 



LANE CHAN 



Channelization code indicating restrictive (left-turn 
only, right-turn only) status of each lane 



65 



+J 

u 



a 
a) 
« 

c 
o 

■H 

-P 
■H 

c 

•H 

m 

ai 
a 



o 
^ 

+j 
c 
o 
u 

H 

c 

■H 

CO 



* — I C <N fN 



CN CN i-i O 



Ss 



*T O P* C <N fN 



B" 



"i O PM CN H C 



C » r*« C <N CN 

a o 

f 



• fN (N H O i-i 



s- £ 



2 s 2 



sis 



s ?s 



P 

1 e 



B B 



B . S 



C 

rd 

c 

•H 
CO 



a, 

e 

CO 






c- 


a. 


1 


z. 




Oi 


^ 


& 


& 


2_ 


Cm 




c 


a 


a 


a. ' 




o 


r~ 


o 


r~ 




T 




^r 


<T 






o 


r- 


o 


[■>• 






"* 


in 


a\ 




w 


CN 


<N 


o 


kfl 


w 


El 

s 


in 


oi 


^ 


*r 


en 


o 


IS 


o 


<£> 


£ 


in 


in 


« 


en 


P- 




c 


a 


\c 


o 


vC 


5 




j-) 


o 








fN 


CN 


r- 


f^- 






V£ 






m 








rt 




















— 










































S 


ST 








cu 


£ 






cu 






a 


£ 


E 


t 




r~ 


m 


r* 


m 


§ 


r* 


r*i 


r- 


r-i 


r- 


m 




r» 


ro 


r- 


m 


§ 


T 




<r 




r- J 




cn 




— 




j* 


^ 




t 














































| 














1 














v 


£ 


T 




o 


T 






^ 


t 




^> 


^» 




*s 


§ 


in 




m 




2 


(N 




TT 




T 




E 


in 




in 





g s 

§ -9 S 
a 2 fe 



S S 6, 

- b E ~ - 

S 55 ? g g 



OrHCN^Tinr^co^ 



IT) 



X! 
Eh 



— r~ CN m «r — h CN r^ v it ifl ^ ~» CN n- -<r 



2 a< tT CT* O 2 OOOCCO £ — — i-i -J 

C C H - — -. _ -• C -l — -I rH 

2 2 2 



2 IN 

C ■-! 

z 



66 



Table 16: Definition of Column Headings in "Sign 
and Signal Control Definitions" Report 



Column 
Heading 

NODE 

INTVL 

DURATION 



Definition 



Node number identifying intersection 

Signal interval number 

Duration of signal interval (seconds) ; 
followed by (in parentheses) percent 
of signal cycle length represented by 
this interval. For sign control only 
one interval is presented of duration 
"0". 



OFFSET 



SIGNAL CODES 
FACING 
INDICATED 
APPROACHES 



Offset of beginning of interval from 
reference time (seconds); followed by 
(in parentheses) percent of signal cycle 
length represented by this offset. 

The upstream and downstream node numbers 
defining the approach links to the sub- 
ject intersection are given as column 
headings. The signal codes defining 
the permitted movements for each approach 
during each signal interval are presented 
under the link identification headings. 
A glossary of the signal codes is given at 
the end of this report. 



67 



in 



o o o o 



-3- 



o o o o 



-p 
u 
o 

a 



ro 



CN 



o 


o 


o 


o 


O 

ro 


o 


o 
in 


o 


o 
<* 


o 
o 


o 

in 


o 
o 



c 
o 

•H 

+J 

•H 
P. 
•H 
<+H 
CD 

a 
o 

•H 
M-l 
"4-1 

a 
u 

EH 

c 

•H 
}-' 
CD 
-P 

c 

w 

(L) 
H 

e 



w 

E-" 

2 

H 
I 

PQ 

D 
CO 



in 



ro 



E-t CN 



oo o o o 



vo in o o 



VD 


o 

rH 


o 

rH 


o 


O 


in 


O 
CM 


o 


O 


o 


O 

in 


o 
o 




CD 

rH 

Eh 



o o o o 

o o o o 

o ro <x> ^r 
ro 



ro 



rH 


*x> 


(N 


in 


O 


o 


O 


o 


r- 


00 


00 


00 



rH ON rO ^< m 



68 



Table 18: Definition of Column Headings in "Entering 
Traffic Definition" Report 



Column 
Heading 



Definition 



LINK 



Upstream and downstream node numbers 
which define each entry link. 



TOTAL FLOW 
RATE 



Rate at which vehicles are generated on 
the entry link (vehicles/hour) 



PERCENT BY 
VEHICLE TYPE 



Percentage of TOTAL FLOW RATE allocated 
to each vehicle type 



PERCENT 
VEHICLES 
BY LANE 



Percentage of TO'T'AL FLOW RATE allocated 
to each lane of link. Applies to 
freeway entries only. 



69 



ro m ro 



m m m 



■P 

U 
O 

CD 

o; 

c 
o 

■H 

4-> 

•H 

c 

•H 
M-l 

CD 

a 

e 
<y 
+j 

CO 

>. 

CO 

0) 

o 
c 
to 



> 

a 

CO 



6 
as 
en 



o m m 



o o o 

in in in 

(N (N CM 



o o o 

in in in 

CM CN CN 



CO 

b 

5 

H 

s 



u 

w 

CO 



•* w 



n 



CO 

g 

H 

I- 

co m 



O "* 



83 



o 

in 



o o 



o ro ^ 



H (N n 



rH cm ro 



ro ro ro 











H 










< 05 










0( M 










< 










C O. 










&: 










hH Q 










Oi w 










D hH 










O 04 










U D 










o 










Pm u 










o 










fa 










Oi o 










o 






« 




O Oi 




w 


<c 




J O 




CU 


Q 




o 




^ 


<C 


04 


2 J 




Eh 


PU 


o 
o 


H 2 




Oi 


« 


l-H 


oi S3 


X 


O 


w 




Eh W 


a 


Eh 


J 


Eh 


co a 


< 


U 


CU 


« 


Z Eh 


co 


w 


P-i 


O 


& CO 


CO 


E-" 


o 


K 


O 0, 


o 


H 


Q 


CO 


Q D 


J 


Q 








O 


W 
Q 




*• 






O 


o 


i-H 


ro <vT 




U 









<Ti 



<D 



■H cm ro 



H ni n 



X! 
(0 
Eh 



rH cn ro 



CN CN CM 



1 



70 



Table 20: Definition of Column Headings in "Surveil- 
lance System Definition" Report 



Column 
Heading 

STATION 
NUMBER 



Definition 

Identification of group of detectors 
which comprise a detector "Station" 
for Incident Detection or MOE 
Estimation 



NUMBER 



Sequence number of surveillance de- 
tector within link 



LANE 
TYPE 

LOCATION 



Lane containing detector 

Detector type as defined in "GLOSSARY" 
at end of report 

Distance from upstream node of link 
to upstream end of detector, or 
acquisition point for doppler radar 
(feet) 



LENGTH 



Detector length (feet), does not 
apply for doppler radar 



71 






o 

in 



« 






o 



4-> 
U 


CD 
(23 

C 

-H 
■P 

•H 

c 

■H 

4-1 

CD 
D 

-P 
C 

cu 

•H 

o 



a, 

e 

fd 



8 



u 

I 

en 



o 



o 

CN 




O r-{ CN 



CM 

cu 

r-H 

rd 

Eh 



^ CO I 

cq r- i 



(Nl I 



O 
O 
O 
O 
O 
O 
(N 



72 



Table 22: Definition of Column Headings in "Incident 
Definition" Report 



Column 
Heading 



Definition 



LINK 



Upstream and downstream node numbers 
which define link 



INCIDENT 
CODE BY 
LANE 



Incident code for each lane of 
Link (as described in glossary at 
end of report ) 



UPSTREAM LOC 



Distance from upstream node of link 
to upstream end of incident (feet) 



LENGTH 
AFFECTED 



Length of roadway affected bv incident 
extending downstream from "UPSTREAM 
LOC." (feet) 



TIME OF 
ONSET 



Time that incident begins measured from 
start of simulation (seconds) 



DURATION 



Length of time incident exists (seconds! 



RUBBERNECK 
FACTOR 



Percentage reduction in capacity due 
to rubbernecking 



73 



S 






■H 

w 

Q) 
Q 

■U 
H 
O 

QJ 
Pi 



O 

•H 
+J 
W 
•H 
4-> 
fO 
-P 
W 

>i 
0) 

n 

<u 
cu 

6 
td 



eg 




3 u-1 


O 


in 


in 


tft 


"> CM 


o 




CM 


04 


D OD 

3 r- 


r- 


CO 






3 CO 




in 
m 




10 


T C* 


O 


r- 


in 


in 


-t in 


CO 


CO 


t 


00 


3 r- 


cr\ 


r*> 


in 


CO 


1 ^£> 
«J r- ! 


en 


o 


fN 


m 

o 



CM ^- r-i 



a I 



i 



in m 



I I 



I I 



8 fn 



* 1 : 

2 § 5 

4 H S 

y g 3 



(L) 

s> 



74 



Table 24 : Definition of Column Headings in 

"Sample Freeway Statistical" Report 



Column 
Headings 

L 

LINK 

VEHICLES IN 

VEHICLES CUT 

LANE CHNG 

CURR CONT 
AVG CONT 

VEH-MILES 

VEH-MIN 



SECONDS /VEHICLE 
TOTAL TIME 

MOVE TIME 



DELAY TIME 



M/T 

VEH-MIN/VEH-MILE 
TOTAL 

DELAY 



VOLUME 



DENSITY 



SPEED 



Definition 

Freeway link index 

Upstream and downstream node numbers 
defining link 

Number of vehicles entering link during 
reporting period 

Number of vehicles leaving link during 
reporting period 

Number of lane changes in link during 
reporting period 

Current number of vehicles on link 

Average number of vehicles on link 
during reporting period 

Total distance traveled on link by all 
vehicles during reporting period (miles) 

Total time spent on link by all vehicles 
during reporting period (minutes) 



Total time spent on link per vehicle 

(seconds) 

Ideal time spent on link per vehicle 

assuming all vehicles travel at their 

desired speeds (seconds) 

Excess time spent on link per vehicle 

(seconds) 

DELAY = TOTAL - MOVE 

Ratio of MOVE TIME to TOTAL TIME 



Time to travel one mile at prevailing 
speed (minutes) 

Excess time to travel one mile above that 
required at desired speed (minutes) 

Flow rate of vehicles through link 
(vehicles/lane/hour) 

Concentration of vehicles per unit road- 
way area (vehicles/lane-mile) 

Prevailing space mean speed (miles/hour) 



75 



fl 


g 


R 


8 


8 


H 


R 


fl 


(0 


rj 


d 


to 


fl 


(fl 


u 


U-) 


u_i 


11* 


U4 


l*-i 


U-i 


N 


* 


3 


* 


a 


N 


N 


u 


U] 


m 


tn 


Ul 


in 


w 



c 

Cn 
■H 
W 
(U 
Q 



o o o 



O © -H o o o 



<N 


0> 


U"> 


<N 


00 


o 


o 


a\ 


m 


T 


O 


TT 


m 






OJ 


in 


o 


*T 


—{ 



-p 

o 

a 



cd 

u 

•H 
4J 
W 
•H 
-P 

id 

-p 
co 

a) 
u 

(0 

4-1 
U 

co 

•a 
c 
m 

& 

(d 

CD 



S8 

h-t M 

** fi 
I 

5 Co 



all 

|§! 

<v 5 M 



= z 

25 



n 



si 



"1 ^ 

o \o 


o 




00 


in 




CO 


fN 

m 


ft 


CO 


TT 


r* 


o 


CO 


O 


VO 


*r 


n 



\0 (N in fN to m 



tOtococnm^ooTCOr-i 

C000fNCn^HC0COr-ltOO"\ 



&4 *H ^ ^ *H 



cor-icoinr^o-Htomo 
cftin\or*cocoino^inoY 



■H ^ -* I 



in 


O 


O 


o 


in 


o 


o 


o 


in 


in 


p* 


a 


in 


fN 




30 


n 


1— t 


o 


£ 


00 


CO 


o 

fN 




P* 

CO 


o 


so 
m 


0"* 

m 


r- 




(N 


fN 


*r 


tO 




m 


CO 


m 


fN 


m 


in 


fN 


o 


10 


0> 


\o 


CO 


fN 


^ 


m 


rH 


fN 


m 


■<r 


CO 


m 


to 


CI 


(N 


fN 



rd 

CO 



fN fN -V 



a§ 



[*~ *r ^O 



<N T I 



ID 
fN 






fd 



co ty. o <-i 



76 



Table 26: Definition of Column Headings in 
"Sample Ramp and Surface Statis- 
tical" Report 



Column 
Headings 

L 

LINK 

VEHICLES IN 

VEHICLES OUT 

CURR CO NT 
AVG CONT 

VEH-MILES 

VEH-MIN 

SPEED 



SECONDS/VEHICLE 
TOTAL TIME 

MOVE TIME 



DELAY TIME 



M/T ' 

VEH-MIN/VEH-MILE 
TOTAL 

DELAY 



PERCENT 
QUEUE DELAY 

AVG SAT PCT 



CYCLE FAILURE 



LINK TYPE 



Definition 

Ramp or surface link index 

Upstream and downstream node numbers 

Number of vehicles entering link during 
reporting period 

Number of vehicles leaving link during 
reporting period 

Current number of vehicles on link 

Average number of vehicles on link 
during reporting period 

Total distance traveled on link by all 
vehicles during reporting period (miles) 

Total time spent on link by all vehicles 
during reporting period (minutes) 

Mean speed of all vehicles on link 
during reporting period (miles/hour) 



Total time spent on link per vehicle 
(seconds) 

Ideal time spent on link per vehicle 
assuming all vehicles travel at their 
desired speeds (seconds) 
Excess time spent on link per vehicle 
(seconds) 
DELAY = TOTAL - MOVE 

Ratio of MOVE TIME to TOTAL TIME 



Time to travel one mile at prevailing 
speed (minutes) 

Excess time to travel one mile above that 
required at desired speed (minutes) 

Percentage of delay time spent in 
queue 

Average percentage of link area occupied 
by vehicles during reporting period 

Number of instances during reporting 
period when queue present at start of 
green was not discharged by end of green 

Identifies link as Ramp or Surface type 



77 



c 

0> 
•H 
(0 
0) 
Q 

■M 

M 
O 

a 

OS 

0) 
0) 

a 

CO 

-a 
c 
«d 

>i 

(15 

a) 

K 
C 

o 

•H 
-P 
«J 
-P 
CO 

>i 
(0 

<D 


<1> 

* 

(O 



l B 



S0 t-4 ~4 



>• tfl --H m4 



in •-« *-* 



en oo 



°. g § 8 

m •■< — t -h 

vc C ci r» 

<5 <r> art 

<N p» 4 » 

_• *> cS c* 

o> o> 



Sao IN 



< m 



1: 



a' 

ScN 



S m in 






a a 

V 00 



•m m •-< m « 



S: 



J 00 

l: 






Is S 

a* * 

§8 S 

is 5 



ss a 



P IN 

fe S o 

3 n o 



in >n 



00 CO 

in rt 

Vfl Id 






in g co 

» « CM 

p» m * 






sis 



| la 

S < to 



§1? 
g S £ 



78 



Table 28 



Column 
Headings 

FREEWAY LINK 



Definition of Column Headings in "Sample 
Freeway Station Headway and Speed" Report 



Definition 

Freeway link index followed by upstream 
and downstream nodes in parentheses 



STATION PLACEMENT Distance from upstream node (feet) 



LANE 



MEAN SPEED 



MEAN HEADWAY 



Lane code number (i.e., l->-5 for through 
lanes, 6->9 for auxiliary lanes) 

Time mean speed for all vehicles cross- 
ing station in indicated lane during 
reporting period (miles/hour) 

Mean time between passage of individual 
vehicles for indicated lane during 
reporting period (seconds) 



PERCENT OF TRAFFIC Subheadings beneath this general head- 



AT OR BELOW 
INDICATED SPEED 



ing indicate the upper limit of each 
cell of a cumulative frequency distribu- 
tion of speed (miles/hour) . Under each 
speed value is displayed the number of 
vehicles which have passed the station, 
in the indicated lane, at or below the 
heading speed value, during the report- 
ing period. 



PERCENT OF TRAFFIC Subheadings beneath this general heading 



AT OR BELOW 
INDICATED HEADWAY 



indicate the upper limit of each cell 
of a cumulative frequency distribution 
of headway (seconds) . Under each head- 
way value is displayed the number of 
vehicles which have passed the station, 
in the indicated lane, at or below the 
heading headway value, during the 
reporting period. 



79 



\ \ \ \ 




00 


in 


n 




iH 






W 


2 


Q 


D 


O 


« 


Z 



(saifiNiw) awn. 



>H 2 

< O 
S H 
Q En 

< CJ 
O W 

or; co 



c 
c^ 

•H 
CO 

(1) 

Q 

-p 

O 
i-H 

>1 
>H 
o 
+J 
o 

<D 
•n 

<d 

u 

Eh 

CU 
iH 

U 
•H 

.c 
> 

6 

Sh 

o 
u 

c^ 

En 
O 

hi 
A 



g 

CO 



u 

•H 



80 



W 

H 

D 

> 

ft 
D 
O 
Eh 

O 
U 



K* I ' l*|H 1 1 i l |H M^ ► 1 ^ l i < Mr^ I i" I &t^ ^r* ^r^ 
MH l-M »-M MH l-M M-| |X| M-j t-M M-l l-U i — ■ — • 

§sss§s§ss§ss§ 

oinoinoinomoLnoLno 



+ X! O 



< a *o> 



A 




CD 



VO 



m 



C 

Cn 

■H 
CO 
QJ 

a 

-p 
o 

.H 
ft 

P 

o 
+J 

o 
u 

e 

u 

& 
o 

n 
ft 

Eh 
O 
k1 

ft 



14-1 

o 







(D 






h 






Pu 






e 






(0 






CO 






• • 






CO 


>t 


c 




fC 





Q) 


£ 


•H 


S-l 


T3 


-p 


3 


tO 





tr> 





a; 


•H 


ft 


w 


ft 



(seq.nuTW) auiji, 



81 



c 
«3 



vo r- 

ro • 

00 ON 

rH ro 



o 



r-» vo vo vc 

ro • on r- 

00 (Ti • i-l 

H ro H • 



O 
(N 
CO 



*3< t^ 
CM v 
00 on 



o 

CO 



r- o 

CM • 

00 o 

rH TT 



r-» co 
on r~ 



O 
O 

CN 
rH 

O 
00 
O 



CM rH rH 

CN • O 

r» o • 

rH ^" CM 



h ^ n 

r-» . o 

r^ o • 

H t (N 



o 
o 

CN 



O 
00 
O 



ON « 

t> o 

rH ><* 



H CN 
O VO 



CO ID m VO 

i> • o * 

r- O • rH 

rH ** CM . 



ON 
CM 

Q) 
H 
X! 

crj 
En 



vT> 



H 

n 
I 

H 



O 
— SO 

u <n 



w 



o 

00 



•if? 

Q 
O 

g ° 

W CM 

o 

H 

En 

6 



o 
o 

v£> 



O 
00 



O 
v© 
en 



«r vo 

cn • 

00 ON 

rH ro 



o t 



on on 

rH ro 



r> rH 



cm ro LD 

•<3" • 00 

CN 00 • 

ro rH 



o in h 

ON • 00 

ON 00 • 

rH ro rH 



ro ON H 
00 • ON 
00 ON • 



r- o rr 

LD • ON 

oo o ■ 

iH <tf H 



8$ 

w 

— O 
Q 00 

o 

H 

tf o 

W CN 



H 

I 



> 



o 
vo 

CO 



m co o ro 

CN • CN 00 

00 CN • rH 

rH rO rH • 



LD CM CTl «3< 

O • 00 CN 

CN ON • H 

H ro H • 



VO VO in rH 

^* " 00 CN 

ON 00 • rH 

H ro rH • 



o\ in h w 

00 • CO o> 

ON 00 • H 

rH rO H • 



in m h on 

oo • on r- 

oo on • h 

rH CO rH • 



vo cm «tr in 

in • on r> 

00 O • H 

rH <tf rH • 



o 

CN 



cm oo ro 

VO • on 

00 ON • 

H ro H 



o 

CN 



in on ro r~ 

VO • on r- 

00 ON . H 

rH ro H • 



o 

CM 



o in on 

O • 00 

ON ON • 

H ro H 



o 

CN 



m in o cm 

CN • ON 00 

00 ON • H 

H ro H • 





82 



£& 



c 
&> 

•H 
CO 
(U 

Q 



CT» 
CTi 



o 
a 



c 
o 

•H 
■P 
A3 

e 

•H 
-P 
(0 

w 

w 
o 



t 



o 
n 

r-l 
.Q 



'J CM 



B 


a 
as 


cu 


H 


B 


H 


o 


8 


2 
O 


£ 



2 3 

•5 M 



E-< 
00 
►J 

M 

co 

8 



CM 

z 



b 2 
O 

M 
CO 



Z 



O <H O CM 

« EH 

z 



O 

H 
Eh 

<c 

Eh 
CO 



M 

z 

8 

w 
o 
*c 
Cm 

CO 



gn 



CO w 

u z 



0< 



w 



en 



CD 



Q 

W K in 

ft S 00 

CO co 



D H ft CO 

j Ej ac oo 

O O a, h 

> > 



CO Tf CM VO CM VO 



vO in cm in 



r^ 
t 






oo 



O VO 
in in 



<j\ <* r~ 
in vo vo 



O r-i 

r» r- 



o 



so <* 

vO vo 



CM 

vO 



oo vo in 



vo co ^r^fH<*>fno 






CO 

in 
oo 



co 



vo 
vO 
00 



00 
CO 



co 

00 



co 



ro 



in o 
en m 
co oo 



o 

CO 



co r~ 

CM CM 



CO CTi 



vo m vo 

CM CM CM 



O PI CM 

o <h m 

CO CO 00 



oo m o 

CM CM CO 



co o m 

O ON CO 

m co co 



a hi o 

D Z Cm CO 

J H E 00 

S £ H 



.h in 

vo vo 

a) co 

.H i-H 



00 
00 



CM 

cn 

CO 



CO o 00 
<T> CTi C7i 
CO 00 00 



in o en 
o ^ <*> 
ct» oo co 



■H VO CTi 

in vo vo 
oo co co 



u 

w 

CO 



M 
Eh 



OOOOOOOO 



o o o o o o 

O CM ^ VO 00 O 
CM CM CM CM CM CO 



83 



G 
■H 

w 

Q 



+J 

O 

Q) 
« 

c 
o 

•H 

+J 

o 

<D 

+J 

a) 

Q 

■P 

c 

<D 
T3 
•H 

O 

c 





z 










c 










H 






CO 




E. 


o 




CO 




U 


H 




Eh Q W 

z u 2 




t-i 


Eh 






CO 


< 


W 


W £ O 






Z 


> 


Q 2 o 




z 


H 


< 


h h a 




H 


S 


S 


U fc, d, 

z z 




E-: 


M 


z 


H O Z 


z 


Z 


Eh 


o 


U H 


o 


u 




H 


E> Eh Eh 


H 


u 


Eh 


to 


Eh 


H 


Z 


CO 


H Z Z 


H 


o 


w 


s 


Eh a W 


z 


g; 


Q 


< D Q 


H 


H 


H 


£ 


Eh H M 


Cn 




U 


Z CJ CJ 


w 


o 


z 


o 


w s z 


a 


z 


H 


u 


Eh H H 


W 


w 


S 


> 


Eh fc Q 
Z 2 U 


Q 


s 


< 


o 


w 


g 


W O Z 


u 


h 


Eh 


HUH 



0) 
H 

e 

fd 

CO 



rH 

En 



.h ir> 




(N <n (N r>4 cn 



84 



illustrates the output of the point processing procedures 
which calculate traffic parameter values on a detector 
specific basis. 

To illustrate the MOE estimation and incident detection 
output formats, it is assumed that a series of fully de- 
tectorized (one detector in each lane) stations have been 
defined by the user along the length of the freeway. Table 
30 is an example of the MOE estimation report, produced by 
INCES, for a consecutive pair of such stations. The "VOLUME 
IN" and "VOLUME OUT" values are mean values across all lanes 
at the upstream and downstream stations, respectively. 
"SPACE MEAN SPEED" and "DENSITY" are estimated generated by 
the particular MOE estimation procedure employed. 

Table 31 is a sample of the incident detection algo- 
rithm output report. The incident history for the simula- 
tion is given as well as the various incident detection al- 
gorithm results. As for the MOE estimation report (Table 
30), one such table is generated for each consecutive pair 
of fully detectorized freeway stations. The sample output 
represents a case where no incident was present in the road- 
way between the detector stations. This is indicated by 
the '"CLEAR" code which appears for every evaluation period 
in the row denoted as "SIMULATED INCIDENT". In the sample, 
Algorithms 1 and 3 correctly deduce that no incident is 
present. Algorithm 2, however, detects a (nonexistent) 
incident commenri na during evaluation period 5 ?.nd ending 
during evaluation period 7. 

3.2.2.5 SAM Module Output Reports 

Reports will be generated by the 
SAM module containing simple comparisons and statistical 
tests of MOE values from separate simulation runs. The MOE's 
to be studied are: 

MOE Application 

Vehicles Discharged All Links and Network 

Delay Time (veh-min) All Links and Network 

Lane Changes Freeway Links 

Density Freeway Links and Network 

Average Saturation Percent Non-Freeway Links 

Vehicle-Miles All Links and Network 

Travel Time (veh-min) All Links and Network 

Volume Freeway Links 

Stopped Delay Time (veh-min) Non-Freeway Links 

Average Speed All Links and Network 

85 



Tables 32 through 36 illustrate the various SAM report 
formats comparing two statistical data sets for a simple 
three link network simulated for three 5-minute time 
periods (subintervals) . The simple data comparisons of 
Table 32 are repeated for all appropriate MOE * s of the 
Freeway and Non-Freeway link categories. The same is true 
for the statistical test results of Tables 34 and 35. The 
term NETWORK may indicate all links, or optionally, all 
freeway links. 

' 3.3 INTRAS Storage Array Methodology 

To cope with a stringent data storage problem, 
most of the INTRAS arrays were designed to contain several 
data items per unit (word/half-word) of storage. This 
"packing" of data is made possible by defining all program 
data elements as integers. For example, the link descrip- 
tors, "number of lanes" and "link length" might be packed 
into one storage element of five decimal digit length, as 
XXXXY; where, XXXX and Y represent the digit positions 
assigned for "link length" and "number of lanes", respec- 
tively. 

Additinally, two procedures were adopted to reduce the 
required storage, both for the arrays, and for the logic to 
access them. First, a dynamic storage allocation procedure 
was developed to ensure that unused space in some arrays 
could be reallocated for other uses. Second , efficient 
packing and unpacking procedures were developed and coded 
into modular subroutines, thereby reducing the extent of 
all logic requiring link and vehicle parameters. 

3.3.1 Dynamic INTRAS Array Allocation 

The largest allocations of computer memory 
in INTRAS are for link-specific and vehicle-specific data 
elements. Since the characteristics of surface, ramp and 
freeway links differ greatly, it was propitious to define 
three different storage structures to minimize memory 
allocation for each individual link and vehicle. It was 
recognized that this procedure could be wasteful in that the 
full memory allocated for the individual arrays may not be 
required for a particular application. To prevent such 
waste, the size of individual arrays are revised within 
the global allocation of storage, in response to the needs 
of each particular application. 

86 



CN 



CO 

c 
o 

W 
•H 

u 

fd 

I 

u 

o 
u 

■H 
re- 
ft 

-P 
C 

a; 

e 
<u 

H 
W 

rrj 
■P 
rcj 
Q 



en 



oo 



<£> 



IT) 



CO 

H 



CO 



CN 
CO 

a) 

H 

Eh 



en 


O 
00 
CN 


H 
r- 

CN 


C^ 


en 

O 

rH 


CN 


CN 
VD 
CN 


rH 

1 


o 
o 

H 


<N 


o 

ro 


O 

rH 
1 




CM 


CN 
CM 


ID 
CN 


r~ 


en 
o 

H 


CO 

in 

CM 


m 

VO 
CN 


CN 

rH 


in 

CTi 


CN 
O 

en 


in 

rH 


en 

1 


CTi 


rH 


cn 


CN 


00 


n 
o 


in 
en 

<N 


00 
00 
CN 


r» 


CN 
O 


o 

rH 
CO 


o 


m 


rH 

O 



rH r-> ^r m 
in CN CN o 
00 00 



CTi in VO <T\ 
OH I ft 
00 00 • 



u) « o co 

O CN CN CTi 
CTl CTi I • 



< « CQ CQ 

2 2 < < 



< CQ CQ 03 
HgT\ 
S 2 < < 



<C CQ CQ CQ 



2 S 3 < 



87 



c 
o 

en 
■H 
U 
at 

Qa 

e 
o 
u 



i-H r- 



o 

-P 
d) 
Z 



< a 



2 
w 
d) 

e 

at 



0) 



r> 


CO 
fS 


cr. 


i-J 
1 


c 
o 

l-t 


in 


T 


<"M 


H 

rH 




CO 


<-^ 


IN 

o 
1-1 


in 

(N 


CM 


—I 


"3 1 

O 

■-t 


0> 
<N 


T 
!N 


cn 


VO 
<N 


co 
rs 


10 
1 


CO 


««• 


«r 


O 


o 
o 


CO 


CO 

r-l 


o 


o 
o 


fl 1 
fM 


IN 


O 


o 

o 

iH 


r-l 
CM 


■a- 

(N 


iH 


i-l 
O 


in 


<£> 


O 
H 


in 


*T 


i-( 


00 

o 


o 

(N 


c 

CM 


o 


o 

o 




CN 


o 


c 
o 

i-H 


VC 
CM 


r- 
in 

CM 



CO CO O CO 



CO CO o o 



O O "f Ol 



r4 T V 



in ic — ' o 



rO CN rH 03 



^TrHtH H ff N H TfOVinCA 



td 

En 



£ £ £ 



< o a c 



< O D Q 



S < < E £ < < 



E E < < £ E < < £ £ < < 

E9 " 2 



< CQ CQ CD 

E S < < 



88 



fO 
O 
•H 
-P 
W 
•H 
-P 
(0 
-P 
W 



1 


IT. 




W 


►j 




1 


■^ 


T 


K 


£ 


w 


6 




c 




K 


a 


C 


\ 



u 

•H 
M-l 
•H 
O 



v. 



44 
C 
•H 

-P 

< o 

en a 

a) 

Q) & 

a -p 
6 w 



W to 



CO 



£> 
En 



89 



u 

•H 
+J 

w 

•H 

+J 
id 
■p 

o 

■H 
IW 
■H 
O 

<u 
a 
w 



H 


o 


o 
o 


H 






o 


o 


r> 




in 




o 


m 


10 



en 
rs 


CM 

in 

H 


o 


o 


H 

i 


1 



(0 

> 

0) 

-p 
c 

■H 
X! 

W 
+J 

s ^ 
< 

w a 

CD 



04 -p 
6 w 
ra <u 

W Eh 



in i-i o\ 
a> oo r- 

(N CN (N 



ID 

m 

<D 

rH 
XI 

rd 
Eh 



90 



•p 

u 
o 

« 

4-> 
W 
Q) 

E-" 



g§ 



CD 



U 
•H 

w 

•H 

-p 
■p 

O 

+J 
<D 
2 



en 
en 



in 


V 


cc 


T 


00 




t-\ 


<M 


<N 









(0 



oc 




91 



The vehicle array memory allocation is truly dynamic, 
changing during simulation whenever the previous allocation 
would be exceeded for either surface, ramp or freeway vehi- 
cles. The remaining available space is then divided propor- 
tionally among the three vehicle arays. Link array memory 
allocation is performed, during the reading of input 
geometry data, to provide ample memory for the particular 
application. After the needed link storage has been estab- 
lished (at the end of input processing) the link memory 
allocation is reduced to the minimum required. 

In the final INTRAS design a large one dimensional 
array, NLV, is defined to contain all link and vehicle data. 
Six two-dimensional arrays - LNKF, LNKR, LNKS , VF, VR and 
VS, are equivalenced to NLV. Potentially then, all elements 
of NLV may be referenced by any of the six specific arrays. 
Consecutive portions of NLV are allocated to the specific 
arrays in the order given above. The size of each of these 
portions is regulated by internal logic to agree with 
current storage requirements. Thus, if only ten freeway 
links are required for a particular application, only 
10 x N elements of storage are allocated for freeway links, 
where N is the number of elements of storage per link. The 
numeric equivalent of N is required as a system parameter 
for each of the six arrays. In addition, two other array- 
specific parameters are required. They are "number of the 
particular link or vehicle type currently allocated," M, 
and "number of link or vehicle vectors not usable because 
of allocation to other arrays," P. The term "vector" is 
used here to indicate a group of storage elements containing 
all parameters for a particular link or vehicle. 

If the current allocation of storage to one of the 
arrays is to be changed (either enlarged or reduced) the 
corresponding M parameter is internally revised. The offset, 
P, (beginning of usable space) , for all following arrays is 
also revised to indicate the required shift of these arrays 
within the global allocation (NLV) . The allocation, M, of 
some other link or vehicle array must, of course, be 
increased or reduced to compensate for the consequent altera- 
tion in available storage. Lastly, the elements of each of 
the affected arrays are relocated to correspond to the new 
total allocation. The following example should clarify the 
storage method: 



92 



1 


42 


20 


2 


18 


20 


3 


22 


30 


4 


22 


300 


5 


8 


100 


6 


8 


— 



The six arrays and their parameters are dpfinprl 
as follows: 

Sequence , 
Array C ontent i Ni Mi 

LNKF Freeway Link Parameters 

LNKR Ramp Link Parameters 

LNKS Surface Link Parameters 

VF Freeway Vehicle Parameters 

VR Ramp Vehicle Parameters 

VS Surface Vehicles Parameters 

In this example, the parameters N and M are as de- 
fined in the precedinq text. M g is not specified, as 
the allocation for surface vehicle will consist of the 
remainder of the NLV array. The problem posed is 
how to assiqn values to P . , such that the storaqe 
elements N^ x M^ required 1 for each array are not 
mutually conflictinq. 

The first array, LNKF, is defined to start at the 
beqinninq of the NLV array. Therefore, its "offset", 
P^ is zero. That is, there are no storaqe elements 
located between the beqinninqs of NLV and LNKF dedica- 
ted to other arrays. 

The first N-. x M-, , or 840, elements of NLV are, 
therefore, dedicated to the LNKF array. The lowest 
numbered element of NLV available to LNKR is 841. 

Dividinq total dedicated storaqe by N2 and 
roundinq all fractions upward yields 



P = 840 = 46 + 12 = 47 
2 18 18 

The first 47 link vectors of LNKR are unusable as 
they are either partially, or totally, devoted to prior 
arrays (in this case only LNKF) in the array sequence. 
Storaqe and retrieval of parameters pertaininq to the 
first ramp link must be accomplished by addinq the 
scalar ("offset") 47 to the ramo link number before 
referencinq the LNKR array. This process is auto- 
matically applied by the packinq and unpackinq routines 
described in the next section. 



93 



The calculation of this and subsequent P- elements 
may be performed via the equation 

p _ < M i-l +p i-l> x N i-1 

i 

remembering to round fractions up. 

Applying this rule, the following results are 
obtained for P.: 

P = (20+47) x 18 _ ,, 

3 22 

p = (30+55) x 22 = 85 

4 22 

p. = m°mi x_22 = 1059 

8 

P = (100+1059) x 8 = il59 
6 8 



Resolution of the final M ± parameter, M g , is accom- 
plished by allocation of the remainder of NLV, where 
the number of elements in NLV is given by R 



M fi = JL - P 



For this calculation, the fraction must be 
truncated. If R = 15000, then 



M 6 = 15000 _ il59 = 716 



Calculation of the H and P parameters is performed 
internally by the INTRAS program on the basis of 
need. During the geometric data input process, a 
reallocation of link space occurs whenever an excess 
of one link category is recognized. A reduction of 



94 



allocated link space- to the minimum necessary, is 
performed after the c ta input process, to provide 
maximum storage for vehicles. During simulation, 
vehicle array space is reallocated whenever the 
number of one categorv of vehicles equals the 
allocated maximum. 

Suppose that at some point in the simulation 
process storage were required for an additional 
50 freeway vehicles. Then; 



M^ = M 4 + 50 = 350 
and, by use of the preceding equations, 



P' = (350 + 85) x 22 _ 1197 
5 8 



P ' = (100 + 1197) x 8 = 12 97 
8 



m; = 15000 _ 1297 = 578 

6 8 



The contents of the VR and VS arrays would then 
be shifted to new areas in NLV corresponding to 
the changes in the corresponding offsets P^ and PX. 
Although this example sacrifices storage originally 
allocated to surface vehicles (i.e., Mg > Mg) , 
the actual program procedure would apply this reduc- 
tion proportionally to both ramp and surface 
vehicle storage allocations. 

3.3.2 INTRAS Data Array Packing and Unpacking 
Procedures 

To simplify the process of accessing link 
and vehicle data to reduce the program debugging activity 
and to reduce the storage devoted to in-line data access, 
modular subroutines are included in the INTRAS design to 
perform parameter packing and unpacking. Two basic pack/ 
unpack activities may be performed. T he first, via 



95 



subroutines UNPAK and PAK, either retrieves or stores the 
entire parameter vector for a specified vehicle, or link, 
of any type. The second activity retrieves or stores one 
selected parameter of the specified vector. 

Whereas PAK and UNPAK may act on any of the six arrays, 
the individual element access routines are array specific. 
This avoids burdening single parameter access with the 
overhead associated with array selection. The eight sub- 
routines which perform single parameter data access for 
ramp and surface link and vehicle arrays are identified in 
Table 2. Freeway link and vehicle arrays contain one para- 
meter per element and are accessed directly to conserve 
time. 

The packing and unpacking process is achieved by lo- 
cating the specified link or vehicle vector in NLV and 
then applying a parameter storage pattern to each element. 
The storage pattern for every parameter of the six arrays 
is a three-digit number, XXY, where 

X = Element of vector containing this parameter, and 

Y = Digit position of element containing low order 
digit of parameter. 

The Y field also implies the high order digit of the 
previous parameter. Parameter storage patterns are stored 
in a separate array which is indexed by unique parameter 
numbers. For example, the following parameter identifica- 
tions prevail for the LNKS array. 

Parameter Storage Pattern 
Parameter Number Code 

Free-flow speed, fps 

Percent of traffic proceeding 

thru 
Total moving time, halves-of- 

an-hour 

If the contents of the 12th element of a particular 
surface link vector were 3044 then 

Free- flow speed =44 fps, and 

Percent of traffic proceeding thru = 30%. 



96 



27 


121 


28 


123 


29 


131 



The Y field for the "Percentage" parameter is 3. This 
implies that digit 2 (i.e., 2 = 3-1) is the high order 
digit of "Free-flow Speed." Also, since the XX field for 
the "Moving Time" parameter indicates that it is in a 
different element than the "Percentage", then the entire 
high order portion of element 12 (starting with digit 3) 
is devoted to the "Percentage." 

The generalized calling sequence for the data access 
routines is CALL Subroutine Name (L, I, JPARAM) where, 

L = The particular link or vehicle number 
I = Either, parameter number (array 

specific) or, an index which defines 
which array is to be unpacked (packed) 
in the generalized vector routines 
(PAK and UNPAK) as follows: 

Array 



1 


LNKF 


2 


LNKR 


3 


LNKS 


4 


VF 


5 


VR 


6 


VS 



JPARAM = location of unpacked information. This 
is a single cell for the individual 
parameter pack (unpack) operations, 
or an array, for the generalized 
operation . 

The following four examples illustrate the use of the 
INTRAS packing and unpacking procedures. 

1) To unpack the current lane (parameter number 4) 
for surface vehicle M, and store in variable LN . 

Call UVS (M, 4,LN) 

2) To pack current lane for surface vehicle M, from 
variable LN 

Call PVS (M,4,LN) 



97 



3) To unpack all parameters for surface vehicle 
(second argument 1=6) M, and store in the first 
17 elements of array IV. 

Call UNPAK (M,6,IV) 

4) To pack all parameters for vehicle M, from the 
first 17 elements of array IV. 

Call PAK (M,6,IV) 

3.4 INTRAS Error Procedures 

Experience with computer programs containing com- 
plex data verification procedures has shown that the re- 
porting of error conditions often proves costly in terms of 
computer storage due to the storage consumed by FORMAT 
statements. A generalized error message procedure was de- 
signed for INTRAS to alleviate the storage demands of 
diagnostic processing. 

Each recognizable error condition in INTRAS generates 
a standard message which identifies the message code number 
and reports the current value of up to 10 pertinent para- 
meters. 

The following is an example of the INTRAS generalized 
error message format: 

**** ERROR 107, PARAMETERS = 27, 93 , 6 

Because the INTRAS program contains nine major overlay 
modules, each with its own family of reportable error con- 
ditions, the message code numbers are stratified by origin- 
ating module as follows: 

Module Message Codes 

INTRAS 1+ 99 

PORGIS 100+399 

LIS 400+499 

POSPRO 500+549 

FUEL 550+599 

SIFT 600+699 

INCES 700+799 

INPLOT 800+899 

SAM 900+999 

For the above example, Error 107 is generated during 
the execution of the PORGIS Module. 

98 



Full documentation of a computer program normally 
includes detailed descriptions of the conditions which may 
cause each error message. Generally, this is necessary to 
clarify these abbreviated error messages. The INTRAS 
documentation includes detailed descriptions of each mod- 
ule's error messages. The following is a group of PORGIS 
Module error messages which might be generated by a typical 
run. The notation P. indicates the appropriate placement 
in the error message for the value of the i parameter. 

107 - Link (Pi, P2) on card type P 3 could not be found in 
link array. Link ignored. Simulation inhibited. 

128 - Upstream node, Pj, of link whose through traffic 
opposes left-turning traffic from link (P 2 , P 3 ) 
specified after first interval; subinterval = P u . 
Card Type 6. Simulation inhibited. 

150 - Link (P )f P 2 ) has designated P 3 on Card Type 4 as the 
lane receiving its lane 1 traffic in link (P : , P„) 
but (P 2 , P„) does not have a lane P,. Simulation 
inhibited. 

169 - Entry in SIGI array for node P,, but no Type 10 card 
was input for this node. Simulation inhibited. 

255 - Incident code specified for non-existent lane P 3 on 
link (P,, P 2 ). Card Type 30. 

P 3 _< 5 - lane number 
P 3 = 6 - 1st left auxiliary lane 
P3 = 7 - 2nd left auxiliary lane 
P 3 = 8 - 1st right auxiliary lane 
P 3 ' = 9 - 2nd right auxiliary lane 
P 4 = - actual number of lanes on link 
P 5 ,P 6 - actual 1st and 2nd auxiliary lanes on 
link 

Simulation inhibited. 

For the example cited above, a link (denoted by up- 
stream node 27 and downstream node 93) is referenced on .* 
surface Link Operation Card (Type 6). This link, was net 
previously defined on a Type 2 Link Geomotrv car;i. 

99 



4. INTRAS FREEWAY PARAMETER CALIBRATION 

UTCS-1 model applications perform as would "real" urban 
traffic networks by virtue of the UTCS-1 logic and a family 
of independent calibration parameters. Extensive evalua- 
tions were performed during UTCS-1 model development to 
measure these parameters and the results of those calibra- 
tion activities are "built-in" to the model. These same 
parameters are imbedded in the INTRAS model to govern traf- 
fic performance on the surface and ramp links. 

Surface street traffic performance is dominated by the 
intersection and intersection dependent phenomena. Accurate 
representation of freeway traffic depends upon the quality 
of representation of vehicular dynamics and interaction in 
an uninterrupted flow scenario. Vehicle interactions are 
represented in INTRAS by the car-following and lane-chang- 
ing logic discussed in the previous sections. The impor- 
tance of vehicle dynamics dictates a more detailed calibra- 
tion of vehicle capabilities and driver characteristics 
than that required for the surface links. 

This section describes the parameters imbedded in the 
INTRAS model to represent the dynamics of freeway traffic 
flow. 

4.1 Vehicle Type Specific Calibration Parameters 

INTRAS is capable of representing up to five in- 
dependent vehicle types. By virture of the capability of 
updating imbedded calibration parameters via card input 
(see Section 2) virtually any desired vehicle type may be 
simulated. A choice was made as to those types of vehicles 
most representative of (without further user input) a 
typical freeway traffic stream. 

This initial selection includes: low performance pas- 
senger car, high performance passenger car, intercity bus, 
single unit truck (of more than four tires) , and trailer 
truck combinations. The single unit truck category speci- 
fically excludes pickup trucks and light vans which are 
similar in performance to low performance passenger cars. 

A survey was performed of prior research on evaluation 
of vehicle performance parameters. Survey results indicat- 
ed that many characteristics influenced the desired ve- 
hicle-type specific parameters, including frontal area, 
weight, power and gearing. It would be possible to define 

100 



Lbs/br 
Horsep 


ake 
ower 


Vehicle 
Length* (ft.) 


< 20 


20 


>20 




20 


100 




43 


100 




26 


300 




53 



vehicle types stratified by each of these parameters. Since 
it is not reasonable to assume that the potential user 
would be able to categorize incoming traffic by these para- 
meters, a practical alternative was chosen. The five ve- 
hicle types described above were identified, and then para- 
meter evaluation was attempted. 

Several sources (Refs. 5-10) proved useful in identify- 
ing vehicle performance characteristics. From these 
sources, one descriptive characteristic, weight to horse- 
power ratio, was identified as having the best correlation 
with performance. The values of this ratio, selected as 
representative of the chosen vehicle types, and correspond- 
ing vehicle lengths, are as follows: 

Vehicle Type 

Low Performance Passenger Car 
High Performance Passenger Car 
Intercity Bus 
Heavy Single Unit Truck 
Truck Trailer Combination 

*The vehicle lengths include a three foot buffer for 
stopped separation. 

Vehicle length characteristics were taken from Ref. 5 
for the 50-percentile vehicle of each class. Truck-trailer 
combinations assumed the mean length of 40- and 50-foot 
wheelbase vehicles. An exception to this procedure was 
made for the bus category. Here, it was felt that the 
General Motors specifications (Ref. 10) were more appropri- 
ate. 

The vehicle type-specific parameters of acceleration, 
deceleration and maximum speed are all influenced by gra- 
dient as well as the vehicle characteristics. The most 
significant work on grade effects, uncovered during the 
vehicle performance research survey, is a "Modified SAE 
Procedure" described in the NCHRP 3-19 Draft Final Report 
(Ref. 8) . This procedure is a modification of the pre- 
dictive algorithm, published by the Society of Automotive 
Engineers (Ref. 11) , to include the effects of coasting 
during gear shifting. The results of the revised proce- 
dure were compared, during the course of Project 3-19, to 
data drawn from References 12 and 13. These comparisons 
served to further refine the algorithm. The majority of 

101 



the INTRAS bus and truck vehicle parameter calibration is 
based on this procedure, as applied to the weight to horse- 
power ratio for the particular vehicle type. 

The following subsections describe the individual para- 
meter evaluations. 

4.1.1 Limiting Vehicle Speeds 

Vehicles traversing highway sections are 
restricted by a maximum attainable speed related to gradi- 
ent and vehicle characteristics. For the gradient levels 
prevalent on most freeway systems, it is assumed that pas- 
senger car type vehicles are capable of eventually attain- 
ing their "desired speed". As such, no attempt is made to 
evaluate maximum attainable speed for passenger cars. Ac- 
celeration and deceleration characteristics of these ve- 
hicles (vs. grade) are treated in subsequent sections. 

From Reference 8, the following limiting speeds (ex- 
pressed in feet per second) are adopted for INTRAS vehicle- 
type calibration. For downgrades it is assumed that all 
desired free-flow speeds may be attained (i.e. are not sub- 
ject to limit) . 
J Grade 

Vehicle Type 

Bus and Heavy Single Unit Trucks 
Truck Trailer Combination 

The source document also describes the deceleration pro- 
files followed by vehicles which enter a gradient section 
at a speed higher than their limiting speed. INTRAS ap- 
plies integer values of both speed and acceleration (de- 
celeration) . The deceleration to maximum speed is best re- 
presented in all cases by a value of one foot per second. 

4.1.2 Vehicle Acceleration Profiles 

Typical vehicle acceleration is influenced 
by speed as well as by grade and vehicle characteristics. 
Therefore, any acceleration calibration must be stratified 
by speed level. In addition, passenger car vehicles are 
not normally operated at full available horsepower during 
acceleration. In this respect, acceleration is a behavior- 
ial phenomenon. Studies have shown (Ref. 6) that target 
speed affects the acceleration profile. The following as- 
sumptions and decisions were made in the calibration of 
INTRAS to cope with these aspects of vehicle acceleration. 

102 



21 


2° 


11 


21 


98 


84 


70 


57 


84 


50 


32 


23 



Speed categories, within which acceleration is 
constant, are defined with a range of 20 feet 
per second each. A definition of normal accelera- 
tion was required for each vehicle type-grade- 
speed category combination. 

Normal mean acceleration for passenger car vehicles, 
at zero grade, are taken from tests performed by the 
Highway Traffic Safety Center, Michigan State Uni- 
versity as presented in Figure 2.9 of the Traffic 
Engineering Handbook (Ref. 6). This source pro- 
vides acceleration profiles for both rural (60 MPH 
target speed) and urban (35 MPH target speeds) 
environments. In INTRAS these two profiles (illus- 
trated in Figures 9 and 10) are used to represent 
freeway and non- freeway zero grade accelerations, 
respectively. 

Variation from the passenger car zero grade mean 
acceleration profiles for high and low performance 
passenger cars is assumed to be proportional to 
the variation in maximum attainable accelerations 
for these vehicle types. Adoption of a combination 
of the "low" and "medium" performance categories 
from Table B-l of Reference 8 as the INTRAS low 
performance vehicle results in an approximate max- 
imum acceleration for this category of 9 feet/ 
second 2 . High performance passenger car vehicles 
are characterized in this source by a maximum ac- 
celeration of 16.5 feet/second 2 . Integer acceler- 
ation rates are chosen, for the INTRAS calibration, 
which ensure that the low and high performance pas- 
senger car vehicles will exhibit respectively lower 
and higher than mean speed versus time character- 
istics during acceleration. To the extent allowed 
by the integer acceleration rate constraint of 
INTRAS, the ratio of high performance to low per- 
formance passenger car acceleration approximates the 
maximum acceleration ratio (16.5 to 9), for each 
speed category. The resulting speed versus time 
profiles are shown in Figures 9 and 10. INTRAS 
calibration passenger car acceleration rates are 
presented in Tables 37 and 38. 



103 



Speed 90 

Feet/ 

Second 




16 20 24 
Time (Seconds) 



Figure 9 



•Mean Freeway Passenger Car Performance 



Low Performance Passenger Car 

High Performance Passenger Car 

( ) Acceleration Rate fpss 

Freeway Passenger Car Zero Grade Ac- 
celeration 



104 



Speed 6a 

Feet/ 

Second 




4 8 12 16 20 24 

Time (Seconds) 
-Mean Passenger Car Performance 



Low Performance Passenger Car 

High Performance Passenger Car 

( ) Acceleration Rate fpss 



Figure 10 • 



Non-Freeway Passenger Car Zero 
Grade Acceleration 



105 



Table 37: INTRAS Calibration Normal Acceleration 

Rates for Low Performance Passenger Cars 



Integer Acceleration Rates in Ft/Sec 2 

Speed (ft/sec ) 















Above 


Grade 


Roadway 


0-*20 


20+40 


40+60 


60+80 


80 


-4% 


Freeway 


8 


8 


8 


5 


3 




Non-Freeway 


5 


4 


3 


3 


3 


0% 


Freeway 


6 


6 


6 


3 


2 




Non-Freeway 


4 


3 


2 


2 


2 


2% 


Freeway 


6 


6 


5 


2 


1 




Non-Freeway 


4 


3 


2 


1 


1 


4% 


Freeway 


5 


5 


3 


1 


i 
1 




Non-Freeway 


3 


3 


1 


1 


1 


6% 


Freeway 


5 


5 


3 


1 


1 




Non-Freeway 


3 


3 


1 


1 


1 


., ... — 


.. . . — - - 


— 





1 





, 



106 



Table 38: INTRAS Calibration Normal Acceleration Rates 
for High Performance Passenger Cars 

Integer Acceleration Rates in Ft/Sec 2 

Speed (ft/sec) 



I Above! 

Grade Roadway 0+20 20+40 40+60 60+80 80 

-4% Freeway -15 14 14 8 5 

5 5 5 

0% Freeway 11 10 5 3 

3 3 3 

2% Freeway 8 4 2 

3 2 2 



Freeway 


•15 


14 


Non-Freeway 


9 


6 


Freeway 


11 


11 


Non-Freeway 


7 


5 


Freeway 


10 


10 


Non-Freeway 


7 


5 


Freeway 


9 


9 


Non-Freeway 


6 


5 


Freeway 


9 


9 


Non-Freeway 


; 6 

1 


4 

i 



4% Freeway 9 9 5 2 1 

2 11 

6% Freeway 9 9 4 2 1 

111 



j 



107 



"> 



• The effect of grade on passenger car acceleration 
as described in the Traffic Engineering Handbook 
(Ref. 6) is based on earlier research by Saal 
(Ref. 14). From graphs presented in this earlier 
source, multiplicative factors were developed 

to relate zero acceleration at each INTRAS grade 
level to that at zero grade. These factors are 
given in Table 39. The resulting accelerations 
are those shown in Tables 37 and 38. 

• Truck and bus accelerations are those predicted 
by the modified SAE truck ability procedures of 
Reference 8. These procedures agree substantially 
with field test data reported by Western Highway 
Institute (Ref. 12), and the Road Research 
Laboratory (Ref. 13). Integer acceleration 
values were chosen for the INTRAS calibration 

so as to provide the closest possible agreement with 
speed-distance curves presented in Appendix E of 
Reference 8. Figures 11 and 12 illustrate the 
resulting speed distance profiles for the selected 
bus and truck types. The values of acceleration 
for each vehicle type, by grade and speed category, 
are presented in Tables 40 and 41. 

4.1.3 Vehicle Deceleration Profiles 

Deceleration rates as applied in the 
freeway logic of INTRAS are determined by the car following 
logic. For the case where deceleration is mandated by the 
position of a leader vehicle, the only necessary calibra- 
tion is the specification of a maximum deceleration, by 
vehicle type. 

Table 2.8 of the 1976 Edition of the Transportation and 
Traffic Engineering Handbook (Ref. 15) contains skidding 
friction coefficients for both new and badly worn tires. 
Application of these coefficients in the stopping distance 

equations : 

V 2 
S = YrTf from Reference 15 



and , 

S = -— — — from the laws of dynamics 



_(!H' 



where 



2d 



108 



Table 39: Multiplicative Factors Relating Passenger 
Car Acceleration on INTRAS Grades to Ac- 
celeration at 0% Grade 



Grade 



Speed (ft/sec) 



0+20 20->40 40+60 60+80 



Above 
80 



-4% 



1.329 1.299 1.539 



1.600 



1.714 



0% 
2% 
4% 
6% 



1.0 



1.0 



1.0 



930 .959 



846 .908 



791 .889 



.833 



.556 



.441 



1.0 



.739 



.430 



343 



1.0 



.522 



109 




m 



- CN 



4-> 

0) 
<U 

M-! 

T5 
C 
(0 
W 

O 
X! 

4-> 



0) 

u 

c 

fO 
4J 
CO 

■H 
D 



Speed (ft/sec) 



w 
^: 
u 

V4 

Eh 

4-> 
■H 
C 
D 



C 

•H 
00 

>1 
> 

<D 

a: 

(T3 

CO 

QJ 
CO 

CQ 

>. 
4J 
•H 

u 

u 
cu 

4-> 

c 

M 

u 
o 

co 
0) 

H 

•H 

o 

c^ 

QJ 

u 

C 

tO 
4-1 

CO 
•H 
Q 

I 
•O 

0) 

0) 

00 



0) 

>4 

•H 
Em 



110 






CD 




CU 




>i 




Eh 




(D 




H 




U 




•H 




Jfl 




0) 




>-^ 




k»' 




M 




U 




3 




5h 




E-< 




5-i 




CD 


, — . 


H 


-p 


•H 


CD 


cti 


CD 


5h 


M-l 


Eh 


K 


5-1 


C 





rC 


IH 


W 




P 


03 


O 


dj 


X! 


rH 


-P 


•H 


— 


m 







CD 


5-1 


U 


CM 


c 




ftf 


0) 


-P 





W 


c 


•H 


fC 


D 


+J 




w 




■H 




Q 

I 




T3 




CD 




CD 




a, 




w 



CM 



CD 
5-1 

•H 
fa 



Speed (ft/sec) 



111 



Table 40: INTRAS Calibration Normal Acceleration Rates 
for Buses and Heavy Single Unit Trucks 



Integer Acceleration Rates in Ft/Sec 2 

Speed (ft/sec) 











, .... . _ 


Above 


Grade 


0+20 


20+40 


40+60 


60+80 


80 


-4% 


3 


2 


2 


2 


1 

i 


0% 


3 


2 


1 


1 


1 


2% 


2 


1 


1 


1 


1 


4% 


1 


1 


1 


1 





6% 


1 


1 


1 









112 



Table 41: INTRAS Calibration Normal Acceleration 
Rates for Trailer Trucks 



Integer Acceleration Rates in Ft/Sec 2 

Speed (ft/sec) 













Above 


Grade 


0+20 


20+40 


40+60 


60+80 


80 


-4% 


2 


2 


2 


1 


1 


0% 


1 


1 


1 


1 


1 

i 


2% 


1 


1 


1 







i 


4% 


1 


1 











6% 


1 















113 



S = stopping distance in feet, 

V = speed of vehicle before braking, mph 

f = friction coefficient, 

and, d = deceleration rate in fpss, 

makes possible the calculation of a maximum deceleration 
rate. For the INTRAS calibration a mean friction coefficient 
of .65 is assumed based upon the values presented in 
Reference 15. The resulting deceleration rate is 21 fpss. 
This value is imbedded in INTRAS for all vehicle types 
except trailer trucks. Figure 2.13 of Reference 6 describes 
the frequency distribution of maximum deceleration for a 
number of vehicle types. The 50th percentile maximum 
deceleration for trailer trucks is 16 fpss in this source. 
This more conservative value is used in the INTRAS calibra- 
tion of the trailer truck vehicle type. 

Deceleration required to respond to signal indications 
and intersection turning movements occurs only on the surface 
and ramp links. The modeling of this phenomena in INTRAS is 
similar to that of the UTCS-1 model. 

Freeway decelerations to achieve a new, lower, desired 
speed are assumed to be "coasting" decelerations (i.e., 
engine engaged, brake not applied). Table 2.5 of the 
Traffic Engineering Handbook expresses passenger car coasting 
deceleration as a function of speed. The following rules 
apply: 

Speed Range Deceleration 
(fps) (fpss) 

0-40 1 

40-60 2 

Above 60 3 

2 
A coasting deceleration of 1 foot/second is used in 

INTRAS for truck and bus vehicles. 



114 



4 .2 Other INTRAS Calibration Parameters 

A statistical basis is required, by INTRAS, for 
the assignment of lane specific volumes and desired free- 
flow speeds. Data, provided by FHWA, representing free- 
flow conditions on the Long Island Expressway, and a wide 
range of volumes on the Los Angeles Freeway system, was 
analyzed to provide the required parameters. The following 
sections describe these activities. 

4.2.1 Lane and Vehicle Specific Desired Speed 

The freeway vehicle generation procedures 
of INTRAS maintain two vehicles in each lane of the freeway 
entry link. As each vehicle is discharged from the entry, 
another is generated. Lane, then, is a prescribed input 
to the vehicle generation procedure. The assignment of 
vehicle-specific desired speed must reflect this a priori 
knowledge. To facilitate assignment of desired speed to 
each new vehicle, it is necessary to develop a relationship 
which defines lane-specific mean desired speed as a func- 
tion of overall mean desired speed. 

The definition of desired speed (also referred to as 
free-flow speed) is that speed which would be maintained 
on a roadway given no impedance from other vehicles (i.e., 
free-flow conditions) . The Long Island Expressway data 
base (reduced from aerial motion pictures) describes 
approximately 125 miles of freeway traffic under free-flow 
(600 to 1000 vehicles/lane/hour) conditions. The subject 
roadway section is three lanes in width and free of ramps. 
The total data base is divided into eight individual sequen- 
ces. If it is assumed that the available data is sufficient 
to describe lane-specific mean free-flow speed for a three- 
lane section, then additional data is still required to 
characterize traffic on roadways of other widths. 

The Los Angeles data base consists of magnetic tapes 
containing raw detector data from all detector stations on 
the San Diego and Santa Monica Freeways. Data for a number 
of fully instrumented (all lanes detectorized) four- and 
five-lane stations have been isolated to provide a basis for 
lane-specific speed evaluation. The chosen stations are not 
in close proximity to either ramps or lane additions or 



115 



drops. A total of 1675 minutes of four-lane data, and 585 
minutes of five-lane data, representative of volumes from 
1000 to 2000 vehicles/lane/hour have been reduced to pro- 
vide the ratio of lane mean speed to overall mean speed. 
Data considered to be drawn from the forced-flow regime 
(i.e., density > 60 vehicles/lane mile, or speed < 40 miles/ 
hour) was specifically eliminated from consideration. Table 
42 presents the resulting speed ratios stratified by 
volume level. 

Since no consistent trend was revealed to indicate a 
predictable variation in the ratio of lane mean speed to 
overall mean speed with volume, it was decided that both 
the 1000 to 1200 vplph and 1200 to 1400 vplph data aggre- 
gates would be employed. Data aggregates for the three-, 
four- and five- lane roadway widths are shown plotted versus 
lateral position in Figure 13. For the three-lane data, 
each point represents one of the eight time sequences of 
the Long Island Expressway data base. One longitudinal 
location (station) was chosen within the extent of the Long 
Island Expressway data base roadway section. Vehicle speeds 
and volumes were reduced from the data base for vehicles 
passing this station. In this way, the statistical data 
reduced from this data base was made comparable to that 
from the detector stations of the Los Angeles data base. 
Also illustrated is a third degree polynomial "fit" to the 
data via a least squares regression procedure. For this 
purpose, the data points were weighted by contributing time 
duration. In addition to the 125 minutes of three-lane data, 
390 and 170 minutes of data for the four- and five-lane 
sections were applied, respectively. 

The resulting INTRAS lane speed calibration consists 
of factors, drawn from the polynomial curve, to be applied 
to overall mean desired speed to generate lane-specific 
values. The final factors are as follows: 



Section 






Lane 






Width 


1 


2 


3 


4 


5 


2 Lanes 


.94 


1.06 


— 


_ 


_ 


3 Lanes 


.93 


1.01 


1.06 


- 


- 


4 Lanes 


.93 


.97 


1.05 


1.05 


- 


5 Lanes 


.94 


.95 


1.01 


1.06 


1.04 



116 



Table 42: Ratio of Lane Speed to Mean Speed for 
Los Angeles Detector Data 



Four-Lane Aggregates 







Volume Category (veh/1 


ane/hour) 




Lane 


1000+1200 


1200+1400 


1400+1600 


1600+1800 


1800+2000 


1* 


.92 


.92 


.93 


.92 


.93 


2 


.99 


1.00 


.97 


.97 


.96 


2 


.99 


1.00 


.97 


.97 


.96 


3 


1.05 


1.06 


1.08 


1.07 


1.07 


4 


1.04 


1.02 


1.02 


1.03 


1.04 



.99 



.93 



.98 



1.05 



1.04 



Five-Lane Aggregates 

.97 .96 .92 



.92 



.97 



1.07 



1.07 



.92 



.98 



1.07 



1.08 



.92 



1.01 



1.04 



1.11 



.86 

.95 

1.00 

1.07 

1.12 



* Note: Lane 1 is extreme right lane. 



117 



Least-Squares Polynomial 



S = .970-.516L+1.889L 2 
-1.376L 3 




Key 



♦ 3 Lane Data 
+ 4 Lane Data 
O 5 Lane Data 



r 

1.0 



l 

.8 



.6 



i 

.4 



r 

.2 



r 1.15 

Ratio, S, 
of Mean 
1.10 Lane Speed 
to Mean 
Speed 

- 1.05 



- 1.00 



- .95 



- .90 



.85 



Left 

Edge 

of 

Roadway 



Lateral Position, L 



Right 
Edge 
of 
Roadway 



Figure 13: Lane-Specific Mean Speed Ratio 



118 



To generate desired speeds for individual vehicles from 
lane-specific mean desired speed, a cumulative frequency 
distribution is required describing variation about the 
mean lane value. 

Individual vehicle speeds for the Long Island Express- 
way data base were grouped into lane-specific distributions. 
One decile distribution was generated for each lane of each 
time sequence. The ratio of cell speed to lane speed was 
calculated for each decile distribution element. A cumula- 
tive frequency diagram of these data points is presented in 
Figure 14. Also shown is a mean curve, based on the data, 
which is used to evaluate the required INTRAS decile distri- 
bution. The resulting distribution, referenced by driver 
type, is as follows: 

Driver Type 12 3456789 10 

% of Mean 

Lane Speed 82 91 94 97 99 101 103 106 109 118 

4.2.2 Lane-Specific Volume Distribution 

In parallel with the determination of lane 
mean speed ratios, the relationship of lane volume and over- 
all volume was calculated. The resulting ratios (lane 
volume/overall volume per lane) are presented in Table 32 
stratified by road width and overall volume level. This 
data indicates that volume level alone is not an adequate 
predictor of lateral volume distribution. Apparently each 
individual roadway displays its own distributive character- 
sties . 

The following summarizes the data sources for Table 32: 

Number Roadway Number of 

of Lanes Identification Stations 

3 Long Island Expressway 1 

4 San Diego Freeway 6 

5 Santa Monica Freeway 3 

As previously mentioned, the stations enumerated above 
were selected for the absence of local geometric effects 
(ramps, lane drops) . Other roadway, traffic and environmen- 
tal (sight lines, truck percentage, urban vs. suburbanite.) 

119 



.5 .6 .7 .! 
100-, L 1 1 1 



-? x -? V W. bl l± ^ ]± ]1 l J 




90- 



80- 



70- 



w 






60- 


<3 


■M 












(U 




O 




H 




(U 




ft. 




a) 


50 


> 




■H 




Jj 




13 




rH 




3 








c: 












SJ 


4 0- 




30- 



20- 



10- 



7** 'vp •'•,rm 



.'5 J 6 A }s ' 



9 1/0 l.'l l. ! 2 1J3 1/4 lJi i]g 1J7 1.3 

Vehicle Speed 
Mean Lane Speed 



Figure 14: Cumulative Frequency of Speed to Mean 
Lane Speed Ratio 



120 



Table 43: Ratio of Lane Volume to Overall 
Per Lane Volume 



Overall 






Lane 






Minutes 


Volume 


Right 








Left 


of 


Veh/Hr/Lane 


1 


2 


3 


4 


5 


Data 


908 


1.30 


1.12 


.58 




_ 


27" 




926 


1.26 


1.14 


.60 


- 


- 


29 




817 


1.36 


1.14 


.50 


- 


- 


12 




1368 


1.38 


1.11 


.52 


- 


- 


16 




818 


1.33 


1.10 


.57 


- 


— 


12 


Long 


657 


1.21 


1.11 


.69 


- 




18 


Island 


741 


1.09 


1.19 


.73 


- 


- 


9 


Express 


813 


1.19 


1.23 


.58 


- 


- 


7_ 


way- 


1118 


.93 


1.05 


1.05 


.97 


— 


105" 




1313 


.92 


1.02 


1.01 


1.05 


- 


285 




1515 


.94 


1.01 


1.01 


1.04 


- 


385 


San 


1702 


.95 


1.00 


1.00 


1.05 


- 


645 


Diego 


1864 


.94 


1.00 


1.01 


1.06 


- 


255 


Freeway 


1117 


.48 


1.00 


1.23 


1.16 


1.13 


25" 




1322 


.58 


1.05 


1.13 


1.16 


1.08 


145 


Santa 


1511 


.69 


.97 


1.11 


1.12 


1.11 


215 


Monica 


1702 


.91 


.91 


.99 


1.06 


1.12 


95 


Freeway 


1909 


1.02 


.91 


.95 


1.01 


1.12 


105 





121 



factors not present in the data base must be assumed to 
dictate lane choice. 

Because no suitable predictor may be developed from the 
available data, the default lane-specific distribution of 
volume in INTRAS is rectangular. This procedure is parti- 
cularly appropriate at or near capacity volume. Since the 
user may well have more specific information, the distribu- 
tion of total volume by lane may be input (see Section 3) . 

Data on the lateral distribution of commercial vehicles 
was taken from two sources. The weaving section trajectory 
data base, acquired from the Polytechnic Institute of New 
York (PINY) for component model validation (see Volume 2), 
identifies each trajectory by vehicle type. Commercial 
vehicle counts were reduced, from this source, for two-and 
three-lane sections. 

Freeway traffic films, taken of the North Central Ex- 
pressway in Dallas (Ref. 15), were reduced to provide further 
data on commercial vehicle lane placement on two-lane sec- 
tions. The commercial vehicle counts obtained from these 
sources are as follows: 

Commercial Vehicle Counts 

Source Right Lane 2nd Lane 3rd Lane 

PINY Experiment 6 
PINY Experiment 7 
Dallas Films 

Figure 15 displays the cumulative frequency of vehicles 
by lateral position based on the above data. The alignment 
of data points permits a straight line approximation (also 
shown in the figure) . This approximation is used to gener- 
ate the following lane assignments for commercial vehicles: 

% of Commercial Vehicles Assigned to Lane 



Right Lane 


2nd Lane 


96 


89 


142 


25 


180 


71 



Number 


Right 


2nd 


3rd 


4 th 


5th 


of Lanes 


Lane 


Lane 


Lane 


Lane 


Lane 


2 


75 


25 


— 


— 


— 


3 


50 


50 





- 


- 


4 


38 


37 


25 





- 


5 


30 


30 


30 . 


10 






122 



100 
90 

80 H 



+j 70 
c 

0) 

o 
m 
a) 60 

Pn 

CD 
> 

•h 50 
-P 

(0 



3 
g 
p 
u 



40 



30 _ 



20 - 



10 - 



r 

0. 

Left 
Edge 

of 
Roadway- 



Key 
+ Two Lane Data 

© Three Lane Data 



20 



.40 



.60 



80 



Lateral Position L 



1.00 

Right 

Edge 

of 

Roadway 



Figure 15: Percent of Commercial Vehicle Population 
to the Left of Lateral Position L 



123 



The basis for the above commercial vehicle lane assign- 
ment approximation is admittedly weak. A data collection 
effort might be undertaken to examine lane assignment by 
volume level, number of lanes and specific vehicle type, 
at locations isolated from on and off-ramp effects. This 
information is not present in the available data bases. 
Such a data collection is beyond the scope of the subject 
project. 

5. LITERATURE REVIEW 

This review is divided into three segments. The 
narrative discussion which follows describes car- following 
traffic models and their application to freeway simula- 
tion. Both car- following models and traffic simulation 
models are reviewed in the form of an annotated biblio- 
graphy of selected publications provided as Appendix C to 
the report. A more extensive Bibliography is also pro- 
vided. 

5. 1 Analytical Car-Following Models 

Initially, a distinction must be made between 
that set of "microscopic" models describing the behavioral 
response mechanisms of individual vehicles, and those mac- 
roscopic" models describing the overall behavior of the 
traffic stream. The work presented here deals with the 
first group of models and is referred to as "car- following 
theory. " 

For the class of microscopic models, a single car is 
assumed to follow a leading car and both are constrained 
to remain in the same lane. The model can be postulated 
as follows: 

Consider a car, n, at a position, x n (t). At time, t, 
its speed is x n (t) and its acceleration is x n (t). A 
trailing car, n+1, is at position, x n +i(t), with speed, 
Xn+l(t)f an< 3 acceleration x n+ i(t). The acceleration of the 
(n+l)th vehicle at time, t, can be expected to depend on 
the relative speeds and separation distance of the two 
vehicles. 



124 



Two primary concerns must be kept in mind. 

• Stability 

• Acceptable realism. 

The question of stability addresses the form of the model. 
Consider the following forms: 

(a) Xn+l^ = l-'txn^) " x n+l^ fc Jl 

(b) x n+1 (t) = K[x n (t) - x n+1 (t)] 

where the parameters, y and K, must be deter- 
mined from field observation 

(c) A linear combination of the previous two laws: 
^n+lft) = y[x n (t)-x n+1 (t)]+K[x n (t)-x n+1 (t)]. 

All these laws are linear laws, which might be appro- 
priate only for small deviations from the desired state 
of traffic. The response of the (n+l)th driver is propor- 
tional to a deviation for which he wishes to compensate. 
The parameters, y and K, are called sensitivities of the 
response to the deviations. Large values of y and K 
correspond to strong compensation, and small values 
correspond to weak compensations. 

A standard approach to investigate the effect of dis- 
turbances and of the stability of linear systems is to 
perform a harmonic (frequency) analysis of the disturbance 
to determine how individual frequency components are 
propagated through the system. Assuming that the deviation 
of the motion of the lead car in a platoon is the source 
of the disturbance, its motion can then be harmonically 
analyzed. When this is done in law (a) , it turns out that 
a resonance exists at frequency, w=y/2. That is, any 
frequency components at frequencies near y/2 are amplified 
strongly by the traffic, the law of amplification of the 
amplitude of the frequency component being [1+w 2 /y]~ m . On 
the other hand, law (b) damps out a disturbance as 
[l+w 2 /K 2 ]~ m for the mth car behind the source of the dis- 
turbance. Hence, law (b) is a reasonable one to investi- 
gate further while law (a) is not. If one investigates 
mixed laws such as (c) or any other law in which the 
acceleration is proportional to the difference in i fc " 



125 



derivatives of the separation distance between two suc- 
cessive vehicles, he finds resonances (instabilities) in 
those laws which contain terms with even values of i. In- 
asmuch as it is doubtful that a driver could be sensitive 
to third derivatives, one is left with only law (b) as a 
possible one for investigation. 

Since responses can never be instantaneous, law (b) 
should be amended to take into account the time lag be- 
tween the time of the actual development of a disturbance 
and the moment of effective response; therefore, law (b) 
should be modified to read: 

x n+1 (t+A) = K[x n (t)-x n+1 (t)]. 

When time lags are incorporated into linear systems, 
instabilities may result. When time lags are long, there 
should be weak response, K, to insure stability. This 
model is stable only when the condition 2KA<1 is satisfied, 
since a disturbance of unit amplitude is propagated back 
to the m"*-" car so that its amplitude at arrival is equal to 
or less than 






[l+(w 2 /K 2 )/(l-2KA) ] 



-m 



Finally, the basic formula can be extended so that it 
is applicable to cases in which large gaps have formed be- 
tween cars. One possible law is that the acceleration of 
the follower should be inversely proportional to the vehicle 
spacing so that 

x n+1 (t+A) = K[x n (t)-x n+1 (t)]/[x n (t)-x n+1 (t)]. 

If we take this model in its steady state condition; 
i.e., A=0, then it can be integrated to give the macro- 
scopic functional relationships between speed, flow, and 
density. May and Keller (Ref.17) in their paper summarized 
in Appendix C, have given a broad summary of car following 
models with various values for the parameters. They 
demonstrate the macroscopic relationships that these para- 
meter values imply and compare them with known empirical 
results. This gives a reasonable indication of possible 
microscopic models for various flow, speed, density 
regions. Unfortunately, no single model fits the complete 
range although this is understandable. 



126 



5. 2 Fail-Safe Simulated Car Following 

For use in digital micro simulations, the 
analytical car-following models described above have two 
drawbacks, they have been developed for a continuous 
rather than discrete time parameter, and no single model 
is appropriate to all traffic conditions. As a conse- 
quence, so called fail-safe models have been developed. 

A fail-safe car-following model is the process of de- 
termining a vehicle's speed and position given that its 
leader has a speed and position that has already been cal- 
culated for the current time scan. Generally, the output 
of the model is the acceleration of the following vehicle. 
A fail-safe model has two elements. Firstly, there is the 
car- following model which calculates the follower's be- 
havior based on some prescribed desired following distance, 
usually a function of the vehicle's speed. Secondly, there 
is an overriding collision prevention model which is based 
on the following vehicle being able to avoid a collision 
when the leader undergoes its most extreme deceleration 
pattern. 

The PITT model described in Appendix B to this report 
illustrates the rationale and development of a fail-safe 
simulation car-following model. 

6. SIMULATION DEVELOPMENT 

6. 1 The Car-Following Algorithm 

6.1.1 Initial Selection 

Five algorithms were considered for possible 
use. A preliminary study of one of these, the Midwest 
Research Institute algorithm (Ref .18) , indicated that it was 
inappropriate for use in the context of the current develop- 
ments, so it was not considered further. The main draw- 
backs of the MRI model were: the algorithm was complicated, 
it had many parameters requiring calibration, and it 
required a long run time on the computer. 

Four candidate algorithms were analyzed in detail. 

(a) The Northwestern Alcrorithm, 



127 



(b) The UTCS-1 Algorithm, 

(c) The Aerospace Algorithm, 

(d) The PITT Algorithm. 

Algorithms a, b, and d are fail-safe types as de- 
scribed in the literature review, while algorithm c uses 
the May-Keller calibration of the analytic type car-follow- 
ing model. The notation used is: 

a = acceleration of follower, 
x = position of leader, 
y = position of follower, 
u = speed of leader 
v = speed of follower, 
L = length of leader, 
T = simulation scanning interval, 

e = maximum emergency deceleration for all vehicles, 
b,k = calibration constants. 

All units are in feet and seconds. 

The Northwestern Algorithm 

This algorithm was developed at Northwestern University 
by a member of the Pittsburgh research team which has, 
therefore, a detailed knowledge of the model and its 
limitations. The model has two key components, a car- 
following rule that sets a minimum following distance 
directly proportional to the following vehicle's speed, and 
an overriding equation that prevents the minimum following 
distance from being violated during times of maximum de- 
celeration by the leading vehicle. 

Given the speed, and location of the leading vehicle at 
the end of the scanning period, the algorithm outputs the 
new speed and position of the following vehicle. The 
mathematics of the algorithm are lengthy and are reported 
fully by Worrall and Bullen (Ref. 19). 

The UTCS-1 Algorithm 

This model consists of a spacing algorithm which pro- 
vides for collision avoidance when the leading vehicle de- 
celerates suddenly to a stop. There is no specific car- 
following algorithm apart from the critical headway calcu- 



128 



lation. The output of the algorithm is given by 

a= [7(x-y-vT-L) + (2u 2 - 3v 2 ) /6] / (v+3) . 

The Aerospace Algorithm 

This model uses the May-Keller calibration of the con- 
ventional analytical car-following model 

a = Av(u-v) /(x-y) 3 

where A is the driver sensitivity factor. 

When (u-v) is positive or close to zero, the above 
formula is inoperative and normal acceleration patterns are 
followed subject to safe spacing limitations. These latter 
relationships are not clearly stated in the paper by 
Harju (Ref . 20) . 

The PITT Algorithm 

This model is founded on a combination of the North- 
western car-following and the UTCS-1 collision avoidance 
procedures. The primary car- following relationship is that 
a following vehicle will attempt to maintain a space 
headway of L+kv+10 feet. The factor, k, which is a func- 
tion of driver type, regulates maximum lane capacity since 
it determines the average headway at high volumes. This 
factor, k, therefore, is also used to establish bottleneck 
conditions since a reduction in lane capacity can be 
achieved through an increase in k. 

The car- following formula is 

a = 2 [x-y-L-10-(k+T)v-bk(u-v) 2 ]/(T 2 +2kT) 

A lag, c, is introduced into the car-following calcu- 
lations after a has been calculated. The lag is applied to 
the calculations of the following vehicles speed and posi- 
tion as shown in Appendix B. Note that c (which must al- 
ways be less than T) is contained explicitly in the 
collision avoidance equations outlined below. 

Overriding this car- following relationship is a colli- 
sion avoidance set of equations which prevent collisions 



129 



when vehicles are undertaking maximum emergency decelera- 
tions. The formula for the emergency constraints are 

a _< - B/2 + [(B 2 +4C)] l/2 /2 

where B = e + 2 (ec+v) /(T-c) 

and C = [2e/(T-c) 2 ] • [x-y-vT-L-cv- (v 2 -u 2 ) /2e] 

provided a > [ (u 2 +e 2 c 2 ) 1//2 -ec-v]/ (T-c) > 
or 

a < 2(x-y-vT-L)/(T-c) 2 

provided -v/(T-c) <a < [ (u 2 +e 2 c 2 ) 1 / 2 -ec-v]/ (T-c) 
or 

a _< -v 2 /2(x-y-L) 

provided a < -v/(T-c) . 

The detailed derivation of this model is given as 
Appendix B. 

6.1.2 Initial Testing 

As the first step in the evaluation of the 
above models, qualitative assessments of their general 
character were made. 

Northwestern Algorithm 

This model is complicated and lacks flexibility. It 
is extremely difficult to set in modular form, since its 
outputs are position and speed rather than acceleration as 
is the case with the other alternatives. Extensive mathe- 
matical reworking would be needed to include a variable 
time scan and driver and vehicle characteristics. At high 
volumes the vehicles tend asymtopically to a state of uni- 
form speeds and headways. Capacity flow is difficult to 
obtain, therefore, while congested turbulent flow is unob- 
tainable. Bottlenecks would be difficult to implement. 

UTCS-1 Algorithm 

The model is simple and easily modularized. It is, 
however, a collision avoidance algorithm only, and there is 
no internal car-following algorithm to generate congested 



130 



flow. As several simplifying approximations have been made, 
mathematical reworking would be needed to allow for differ- 
ent vehicle and driver types and for variable scanning 
periods. The model cannot reproduce bottlenecks and vari- 
able traffic conditions. These must be imposed exogenously. 

Aerospace Algorithm 

The major problem with this model is that it is a 
part of a completely different simulation design, with the 
basic freeway a set of cells rather than a continuous 
coordinate system. The car-following formula is valid only 
for a positive closing speed between vehicles. For other 
conditions, the algorithm is less clearly defined. As a 
consequence, the performance of the model is uncertain, 
especially with regard to the collision avoidance charac- 
teristics. The reproduction of bottlenecks is not easy and 
the analytical car-following model will be unstable at 
longer scanning intervals. 

PITT Algorithm 

This is a development which combines many advantages 
and eliminates many disadvantages of the Northwestern and 
UTCS-1 models. The PITT model is simple, flexible and 
easily adapted to modular form. The model is mathemati- 
cally rigorous. It easily accommodates variable scanning 
periods, and different driver and vehicle types. Capacity 
conditions can be replicated and congestion is internally 
generated. Bottleneck conditions can be easily imposed 
over the full range of potential capacity reductions. 

Operational Tests 

The four algorithms were each given some initial 
operational tests through simulating the car-following be- 
havior in a single lane. Platoons of two vehicles and five 
vehicles were run down the lane at a constant speed. An 
artificial velocity disturbance was applied to the leading 
vehicle, and the behavior of the followers was examined. 
Scanning times were varied over a range of 0.5 to 5.0 
seconds. Typical outputs of this phase are shown in 
Figures 16 to 20. 

The figures show the behavior of a five-vehicle platoon 

131 



/; 






K 



/ 



<-\ 


co 


in 


3* 


4t= 


=*= 


Q) 


CU 


CU 


H 


rH 


rH 





U 


u 


•H 


•H 


•H 


Xi 


£ 


X! 


CD 


CD 


CU 


> 


> 


> 




o 




rH 


^ 




rd 
> 

0) 

■p 
c 

H 

G 

O 

cu 

CO 


o 






en 




CO 

c 

o 

1 

e 

x\ 

+J 

•H 




^-s 







CO 


tr> 




T3 


H 




c 


<C 











o 


En 


o 


CU 


Eh 


CN 


CO 


H 




' 


Cm 




CD 


.. 




e 


5H 




•H 







rH 


•H 

> 

id 

CU 

C 




o 

rH 






■p 
fd 



Pn 



V£> 



CU 
M 

0-. 

•H 
Cm 



O 



O 



O 



Speed (ft. /sec.) 



132 




o 




rH 


•>* 




CO 

> 

CD 
+J 
C 
H 

-d 

c 



u 

CD 
CO 


o 






m 




0) 

CD 

1 

e 

4-> 




^-s 


•H 




W 


H 




T3 







C 


tP 







rH 




u 


<C 


o 


CD 




<N 


CO 


E-t 




'" w- ' 


E-t 
H 




Q) 


CU 




e 






•H 


• • 




Eh 


u 
o 

•H 

> 

CD 


o 




c 


■H 




o 
o 

+J 



cu 



CD 
•H 



Speed (ft. /sec. ) 



133 




•^r 




> 

c 

H 

C 
O 
O 
0) 

to 


o 




(U 


ro 




c 
o 
1 

g 




J-^ 


•H 




CO 


5-1 




T5 







c 


tn 




o 


iH 




o 


< 




<D 






w 


CO 


o 


*■* 


u 


CM 




En 




0) 


D 




g 






•H 


»• 


o 

iH 


Eh 



•H 
> 

Q) 

m 
c 



4J 

td 






Ph 



00 



0) 

u 

•H 



Speed (ft. /sec) 



134 




o 




fd 


rr 




> 

Q) 
-P 
C 
H 

H3 

C 


Q) 

cn 

0) 


o 




CD 


m 




n 

E-» 
1 

e 
■p 

•H 




- — . 







w , 


Cn 




T3 


f-\ 




S 


< 











u 


CO 


o 


<u 


u 


CM 


en 


E-t 




*-* 


D 




CD 


.. 




g 


S-i 




•H 







En 


•H 
> 

cu 

PQ 

c 


o 







rH 





(0 



CU 



CT> 



CU 

Cs. 
•H 
fa 



Speed (ft. /sec.) 



135 



rH 


m 


in 


=**= 


=«= 


=#= 


<u 


CD 


<u 


rH 


rH 


rH 





U 


U 


•H 


•H 


•H 


£ 


Xi 


x: 


CD 


CD 


cd 


> 


> 


> 



o 



o 




o 



- o 



w 

C 

o 
u 

0) 
CO 



£ 

•H 



o 



o 

«3< 



o 



f0 

> 
u 

CD 
•P 
C 
H 

•d 

e 
o 
o 
CD 

CO 

CD 
C 
O 

I 

e 

-P 

•H 

>-) 
O 

rH 

cu 

O 
CO 

o 

CD 
< 



U 
O 
•H 
> 

fd 
,c 

CD 
CQ 

C 
O 

o 
-p 
fd 
rH 



o 

CD 
U 

CP 
•H 
fr, 



Speed (ft. /sec.) 



136 



traveling at 60 feet/second/second, with either a one- 
second or three-second scanning interval. The velocity of 
the leader was varied by applying an acceleration of -6 
feet/second/second for 6 seconds, a zero acceleration for 
3 seconds and an acceleration of 6 feet/second/second for 
6 seconds. The figures illustrate the velocity response of 
the third and fifth vehicles in the platoon. 

Figures 16 and 17 show the PITT model at one-second and 
three-second scanning intervals, respectively. The results 
are excellent with the following vehicles demonstrating 
good oscillatory behavior, while remaining fundamentally 
stable. The behavior at the longer scanning interval was 
reasonably consistent. Overall, under the simple opera- 
tional tests, the PITT model consistently showed satisfac- 
tory behavior. 

Figures 18 and 19 show the UTCS-1 model at one-second 
and three-second scanning intervals, respectively. The 
results at the one-second interval show reasonable and con- 
sistent behavior, but the speed patterns of the followers 
appear excessively damped without the oscillations demon- 
strated by the PITT model. This damped pattern would 
result in an unsatisfactory representation of congested 
flow. At the three-second interval, the model became most 
erratic with very atypical vehicle behavior. 

Figure 20 shows the Aerospace model for a one-second 
scanning interval. The behavior was very damped and unsat- 
isfactory. A major problem here is that apparently the 
model uses unpublished algorithms for positive accelerations 
and/or zero speeds. Hence, we could not get the vehicles 
out of the low steady state speeds that the system set it- 
self into. For the same reasons, the three-second interval 
was a failure as the following vehicles merely stopped and 
stayed stopped. This has not been diagrammed as it is of 
little interest. 

The Northwestern model has not been diagrammed, as it 
also was of little interest. The car- following behavior 
was even more damped than UTCS-1 at one-second intervals, 
and the model has not been developed for any other interval 
range. 

In summary, these simple tests indicated that the PITT 
model alone shows all the desired characteristics; namely, 



137 



good but not excessive oscillatory following behavior and 
reasonable consistency over a range of scanning intervals. 

6.1.3 Final Selection 

The evaluation of the four models covered 
the following factors. 

Simplicity : All except the Northwestern model meet 
this criteria. 

Internal Consistency : The Aerospace algorithm as 
given, applies only to a restricted range of relative 
speeds. The information as to how it handles the more 
extensive conditions is not published. In the UTCS model, 
basic vehicle behavior is not internally generated, since 
the algorithm is basically an anti-collision rule rather 
than a true car- following rule. The other models are 
good. 

Applicability : The Aerospace model is least satis- 
factory as it was designed for a totally different simula- 
tion approach. All the others are good. 

Variable Time Scans : Only the PITT model is specifi- 
cally designed for this, while the others need reworking. 
The analytical car- following of the Aerospace model will 
be the first to break down as the scanning periods are 
increased. The UTCS-1 model also becomes erratic at longer 
intervals. 

Driver and Vehicle Characteristics : Only the PITT and 
Aerospace models handle this in a generalized manner. 

Capacity Reductions at Bottlenecks : Only the PITT 
model is specifically designed to achieve this which it 
does simply and efficiently. This functional requirement 
is a most important feature of SIFT. 

Operations at the Limits : The Aerospace model requires 
overrides at low speeds and large headways. The North- 
western model is unsatisfactory at high volumes. 



138 



Operational Testing : The PITT model was most satis- 
factory while the Aerospace or Northwestern models were 
the least satisfactory. 

Table 44 summarizes the selection process. 

Under each criteria, the models have been ranked 
according to the test results and the qualitative analyses. 

Summary 

The PITT model was selected as it appears to be equal 
or superior to the others in all categories. The main 
alternative is the UTCS model but most of the key charac- 
teristics of this model have been included in the PITT 
model with an expanded region of application and greater 
mathematical rigor. 

6 . 2 Lane Changing Development 

Careful attention was given to the lane 
changing component, since it is an essential requirement 
that the model satisfactorily perform lane changing and 
merging at high volumes. It is also essential that the 
lane changing component be fully integrated with the car- 
following component. To meet these requirements, an 
algorithm has been developed beyond the fundamental 
approaches used by other simulations although the method 
has many similarities to that used by the Midwest Research 
Institute Simulation (Ref.18). In the usual approach, as shown 
in Figure 21, a vehicle wishing to change to another lane, 
Vehicle 3, looks at the gap available in that lane and 
carries out the following checks: 

1) Does the lead headway to the gap leader, Vehicle 1, 
satisfy the car- following rules? 

2) Does the lag headway to the gap follower, Vehicle 
4, satisfy the car- following rules? 

If the answer to both is yes, then the vehicle can move 
to the new lane. 



139 



Table 44: Ranking of Car-Following Algorithms 



Criteria 




Algori 


thm 








Aerospace 


Northwestern 


PITT 


UTCS-1 


Simplicity 


2 


4 




2 


1 


Consistency 


3 


2 




1 


4 


Applicability 


4 


1 




1 


1 


Variable Scans 


4 


3 




1 


2 


Driver and Vehicles 


2 


4 




1 


3 


Bottlenecks 


3 


3 




1 


2 


Limiting Operations 


3 


3 




1 


1 


Operational Testing 


3 


4 




1 


2 


Summary 


3 


4 




1 


2 



140 



LO 



V 



LA 



ro 



V 



y 



* 



CD 

iH 
U 

-H 

x: 

> 

•H 

Cn 
C 
nd 
X! 
U 

CD 

C 

CD 

xs 



CN 




u 



CM 

CD 

o- 

•H 
En 



141 



With the INTRAS prototype lane changing process, these 
basic checks remain, but the lane change takes place over 
a finite period of time; i.e., the time usually taken for 
a vehicle to physically change lanes. This time is an 
input variable to the simulation. Worrall and Bullen (Ref. 
19) have shown that the time to change lanes is somewhat a 
function of vehicle speed. This relationship, however, 
is not a strong one, and a constant changing time is a 
reasonable assumption as the variation is very small com- 
pared to the normal scanning interval. 

While the lane change is actually in progress, the 
changing vehicle is represented in the simulation as 
occupying the target lane. The original follower (vehicle 
5 in Figure 21) is flagged to identify it for the duration 
of the lane-change maneuver. The model logic can then 
retard acceleration of the flagged follower to fill the 
"vacuum" caused by the lane-change. 

The generation of the decision to lane change has some 
important modifications. Firstly, in determining a safe 
headway, the changing vehicle must satisfy only the non- 
collision constraint equations for the gap in the new lane, 
rather than the car-following equations. This allows 
finer tolerances and expedites lane changing in heavy flow 
conditions. It is considered that these finer tolerances 
are, in fact, present in the actual lane changing process. 
Secondly, the changing vehicle need only occupy this "safe" 
position at the end of the lane change; i.e. a temporary 
unsafe position is allowed during the finite period of 
the lane change. This enables the representation of forced 
lane changing, with a vehicle crowding into what might 
normally be considered an unavailable gap. 

The lane changing logic for the prototype model, 
therefore, consists of the following: 

1) Check if the vehicle must change lanes and flag 
with a lane change desire. This will occur if 
the vehicle is exiting the freeway or if it is 
entering the freeway from an acceleration lane. 

2) Check if the vehicle might want to change lanes 
even though it is not essential. This might occur 



142 



if the vehicle is below its desired speed and 
wishes to change lanes so that it can pass a 
slower vehicle. The desire to do this is 
generated randomly according to a binomal 
probability. This is known as the probability of 
lane change (PLC) and its calibration is des- 
cribed later in this report. The PLC generation 
is applied at user specified frequency (default 
is every second time-step) to all vehicles 
traveling below their desired speed. The desire 
to change lanes might also occur if the vehicle is 
at its desired speed but is seeking to regain its 
desired lane. Again, the PLC generation is 
applied. 

3) To prevent oscillatory lane changing, a few 
additional constraints limit the non-essential 
lane change desire. A vehicle will not change if 
it is accelerating at a rate greater than one 
foot per second. Also, it will not change if the 
potential new leading vehicle in the new lane has 
a speed lower than that of the current leader, and 
at the same time the potential lead headway is 
less than the current lead headway. Also, vehicles 
will tolerate lower speeds before changing to 

the right, and they will avoid the extreme 

through lane adjacent to an acceleration auxiliary. 

4) Vehicles flagged for a lane change are now checked 
to establish whether the change is possible. The 
first step is to check the lead headway in the 
desired lane. The acceleration that the changing 
vehicle must undergo so that it will be safely 
behind the new leader at the end of the lane 
change time is predicted. Two classes of change 
are dealt with, free changes and forced changes. 
For a free change, this acceleration must be 
positive or zero, i.e., the changing vehicle can 
change lanes without lowering its speed. For a 
forced change, which will occur if the changing 
vehicle is making an essential lane change, then 
some decelerations are accpetable and a forced 
change may be instituted. If the lower limit of 
acceptable deceleration, however, is less than 



143 



the minimum normal acceleration (maximum decelera- 
tion) for the vehicle, then no change can be 
implemented. 

5) If the lead check is successful then a check of 
the lag headway is made. The acceleration that 
the following vehicle must undergo so that the 
changing vehicle can safely pull over ahead, is 
predicted. If this acceleration is greater than 
zero, then a free lane change can be started. 

In the case of a forced change, the following vehicle 
may decelerate to allow the changer into the gap. A 
random proportion of followers, "the courtesy factor," are 
assumed to have this characteristic. For a non-courteous 
driver, a forced change cannot be implemented. For the 
courteous driver, the forced change can take place only 
if the calculated decleration is less than the vehicle's 
normal maximum deceleration (minimum acceleration) . 

The lane changing process described above is flexible 
and efficient. It is particularly suitable for simulating 
merging and weaving under very congested traffic conditions 



144 



6 . 3 Vehicle Generation 

Vehicle generation takes place on an entry link 
which feeds the first link of the freeway. The vehicle 
characteristics are randomly generated, i.e., driver type, 
vehicle type, desired lane, and desired speed. Once these 
have been established, the key variables that must be 
determined are the actual speed and position of the vehicle 

Initially, the vehicle is given an actual speed which 
is the lowest of its desired speed or the actual speed of 
the next vehicle ahead. This simple rule substitutes for 
the more accurate value which is very difficult to obtain 
since speed is an independent variable in the car-following 
equations. Next, the time of arrival at the upstream end 
of the first freeway link is predicted for the leader 
vehicle. The prediction method is structured so as to be 
biased towards early arrivals. The new vehicle is placed 
so as to arrive at this boundary at the proper interval 
(dictated by the specified demand volume) after the 
leader's predicted arrival. 

If the early predicted arrival bias results in an 
excess of vehicles being generated, then further vehicle 
generation is inhibited until a balance is achieved. 

The headway is checked through the car-following 
equations and, if too short, is adjusted upward to the 
minimum safe following position. The speed and position of 
the new vehicle are thus determined. 

Vehicles are generated such that each lane of the 
dummy link always has at least two vehicles in it unless 
an excess of vehicles has already been generated. In this 
way, each generated vehicle has time to respond to the car- 
following rules and be operating normally by the time it 
enters the simulated freeway. The simulation is initial- 
ized with an empty system and fill-up must be allowed. 



145 



7. COMPONENT MODEL TESTING 
7. 1 Calibration 

7.1.1 Data Base 

The data bases used for calibration con- 
sisted of general freeway capacity characteristics 
available in the literature including lane capacity, lane 
changing intensities, and ramp merging capacities; the 
Long Island Expressway data set; and the Ohio State vehicle 
trajectories. The Long Island sets have been of limited 
value since there is a scarcity of congested data which is 
the main traffic flow regime in which the calibrations are 
needed. Other data sets used for validation are described 
in Section 7.2. 

7.1.2 Procedure 

The primary calibration procedure has been 
to establish the sensitivity of the simulation outputs to 
the parameters that can be varied. This allowed initial 
parameter specification to provide an operating simulation 
while also allowing users the option of alternative ranges 
of operations if they are prepared to adjust the parameter 
values. 

Car-Following Calibration 

The Ohio State trajectories were used mainly as a 
validation tool, but they were also used as indicators of 
some general car- following characteristics. In particular, 
they indicated the need for the b(u-v) 2 term in the car- 
following equations to correct for anomalies when the speed 
of the follower is very high relative to the speed of the 
leader. A value of b=0 . 1 was calibrated from the data set. 
Further, at very low speeds, the data indicated an inter- 
vehicle spacing of 10 feet which was added to the desired 
headway specification. Also, from this data set, the value 
of the lag, c, was set at 0.3 seconds, and 0.2 seconds for 
negative and positive accelerations, respectively. 

The sensitivity of the simulation to the k factor is 
shown in Figure 22. From known capacity characteristics, 
a mean value for the k factor to give a lane capacity of 



146 



VD 



u 

w 
co 



>H 
EH 

H 

> 

H 
En 
M 
CO 



-H 
U 
fO 

a 
rd 
o 

0) 

c 

<D 

Di 
fC 
5h 
(U 

> 
< 



\M 



CM 




>1 


• 


CO 


■H 


■H 


P4 


> 

•H 




> 






H 






Q 


CD 


00 


2 


CO 


• 


<C 


u 

0) 

> 


O 


w 




s 






•H 






SH 






Q 






## 


^J* 




CM 


• 




CM 


o 




CD 

•H 
fa 



O 


o 


o 


O 


O 


O 


O 


o 


o 


O 


O 


O 


in 


<N 


<T> 


<o 


CO 


O 


<N 


<N 


rH 


rH 


rH 


rH 



AVERAGE LANE CAPACITY 



147 



just over 2000 vehicles per hour was selected. This value 
is k=0.93. 

Lane Changing Calibration 

The key parameters are the courtesy factor and the 
probability of a lane change. The freeway simulation was 
operated for a range of values for the "probability of lane 
change" (PLC) . The resulting lane change intensities are 
shown in Figures 23, 24 and 25 and in Table 45 for 
two-, three- and four-lane roadways. These were then com- 
pared to the empirical results from the Northwestern 
University lane changing report and an appropriate proba- 
bility of lane change of 0.05 assigned. The choice of a 
PLC of 0.05 agrees closely with the data for two- and 
four-lane roadways. For the three-lane case, however, the 
indicated value is 0.10. It should be noted that this 
calibration is dealing with through traffic only, and in 
the Northwestern three-lane case, the Stevenson Expressway 
in Chicago, ramp influenced lane changes probably account 
for the higher frequencies observed. Both the 
two-lane case (the Tri-State Toll Road) and the four-lane 
case (the Dan Ryan express lanes) are relatively free of 
ramp effects. 

The courtesy factor was calibrated by checking the 
capacity performance of a single-lane entrance ramp 
merging with a two-lane freeway. The ramp capacity was 
compared with freeway capacity and Figure 2 6 shows the 
result. 

As the courtesy factor increases from zero, the free- 
way lane capacity drops rapidly at first but soon levels 
out. Similarly, ramp capacity increases initially, but it 
too then levels out. A courtesy factor of 5 per cent was 
chosen as the approximate point where ramp capacity 
slightly exceeded the capacity of the shoulder lane of the 
freeway. 

Vehicle Generation Calibration 

Vehicles are generated using a negative exponential 
gap distribution. Lane volume is determined by the mean 
gap length which is a variable input. The program was run 
for varying values of mean gap length and the output 



148 



o 

CM 



U 

a* 



o 
II 
u 



ID 
O 

* 

o 

II 
u 

04 



o 
o 

CM 




O 
O 
O 
CM 



O 
O 



O 
O 
CM 



> 



QJ 




> 



o 
o 

00 



en 
C 

•H W 



u 

0) 
C 
(0 
h3 



o 

Eh 



(-1 


m m-i 
o 

>i e 

P P 
•H .H 
W 

c > 

d) 

-p m 

c o 

H 


•H 
-P 
O 

c 

p 

Pn 

fd 



n 

CM 

0) 
P 

■H 

fa 



o 
o 



o 



O 



o 

CM 



1^ 

o 



Mean Number of Lane Changes 
(5-Minute Interval) 



149 



o 

CM 

o 

II 



o 
o 



in 
o 

o 

II 



o 
o 

CM 




O 
O 
00 



O 
O 
CM 



fa 
> 



i 

iH 


> 



o 
o 



(0 
& Q) 
C C 
■H fO 
CHh^ 
C 

o u 

£1 
(D Eh 
C 

^1 
M-i 
M-l 

<U 
g 
>i P 
-P .H 
•H 
CO > 
C 
IH 

-P o 
c 

H C 

o 

-H 
.. +j 

^ o 

CM C 

Q) fa 
H 

3 (0 
& 
■H CO 

fa rd 



o 



O 



o 

CM 



Mean Number of Lane Changes 
(5-Minute Interval) 



150 



o 

CN 



O 

rH 

o 



in 
o 




o 
o 

CN 



O 
O 

o 

CN 



O 
O 



O 

o 

CN 



EC 

> 



e 
p 


> 



o 
o 

00 



o 
o 



c cu 

■H C 

C .-3 
fO 
,C 
U 



U 

P 

o 

0) Cm 



C 

o 



u 

o 



>i P 
+) .H 
•H 
W > 
C 

(1) M-l 
-U 



C 

o 

•H 

+J 



c 
p 

En 



in 

CN 



«J 



0) 

P 
5" CO 
' r H rcj 



o 
o 



o 

00 



o 



O 

<3* 



o 

CN 



Mean Number of Lane Changes 
(5-Minute Interval) 



151 



Table 45: Mean Frequency of Lane Changing vs 
Probability of Lane Change and Volume 



4 Lanes Section 

Mean Number of Lane Chanqes in 5 Minutes 

P robability of Lane Change 

Volume 
500 
1000 
1250 
1500 
1750 
2000 

3 Lanes Section 

Mean Number of Lane Chanqes in 5 Minutes 

Probability of Lane Change 



0.02 


0.05 


0.10 


0.20 


9.8 


16.4 


28.0 


45.6 


17.6 


32.6 


51.0 


83.0 


21.6 


41.0 


67.4 


94.6 


19.8 


43.6 


66.4 


93.6 


22.2 


34.4 


59.2 


92.4 


12.0 


32.8 


50.6 


73.4 



Volume 


500 


1000 


1250 


1500 


1750 


2000 



0.02 


0.05 


0.10 


0.20 


6.2 


13.0 


19.6 


26.8 


12.2 


28.0 


39.0 


53.8 


9.8 


29.2 


37.4 


66.0 


15.2 


30.0 


43.2 


60.0 


13.6 


26.0 


39.8 


58.8 


9.8 


22.2 


20.6 


48.6 



152 



Table 45: Mean Frequency of Lane Changing vs 

Probability of Lane Change and Volume (continued) 



2 Lanes Section 

Mean Number of Lane Chanqes in 5 Minutes 



Volume 
500 
1000 
1250 
1500 
1750 
2000 



Probability of Lane Change 



0.02 


0.05 


0.10 


0.20 


4.0 


9.6 


13.6 


17.2 


8.0 


10.4 


19,8 


27.2 


7.4 


13.6 


22.6 


36.6 


10.4 


15.0 


22.2 


40.4 


9.4 


14.2 


23.4 


34.2 


5,0 


12.8 


15.4 


29.0 



153 



o 
o 





0. 


1 








g 


I 






tC 


/ 






« 


1 








I *■"* 


.. 




(L) 


n 






o 


/ Q 






c 


/ *0 






rt 


/ H 






m 


/ 3 






+j 


/ o 






c 


/ -C 






H 


/ W 
/ H 








/ * 


- 






/ ^ 








/ c 








/ (d 








I J 




,*— «* 








G 








rC 








•H 








'O 








<U 








s 






- 










CN 








=*t= 








CD 








c 








fd / 








^ / 






r i 




T 


- ■ 1 


r 







c 









o 




■H 


00 




+1 


• 




(0 


o 




Calibr 
p Nose 




u 


s 







H rd 


o 


■P 


tf 


<£> 


O 


■P 


• 


to 


O -P 


o 


fa 


rcj (0 
fa 




>i 


T5 




CO 


>i <u 




<D 


CO ^ 




+J 


CD 3 




U 


+J w 




3 


U ro 







P CD 


o 


u 


s 


*tf 




u 


• 




>1 


o 




re 26 : 
Capacit 


o 




P 


CN 




Cn 


• 




•H 


o 




fa 



o 
o 



o 

o 
o 

CM 



o 
o 



o 

o 

CN 



O 
O 
CO 



O 
O 



Lane Capacity (VPH) 



154 



volumes recorded. Figure 27 shows the relationship between 
mean gap length and actual generated traffic flow, and this 
is the relationship currently used in calculating volume 
inputs. 

7. 2 Validation 

7.2.1 Data Base 

Data bases that were used for validation of 
the simulation include the Ohio State trajectory data, the 
Long Island Expressway data, the PINY weaving data from the 
Long Island Expressway, and some of the Los Angeles 
closely spaced data set (30 minutes of data for three sets 
of detectors at about 600 feet spacings) . Also, as part 
of the validation process, the general model outputs have 
been studied for consistency, and reproduction of known 
traffic characteristics. The major validation was with the 
PINY weaving data set. Good data was available for the two 
locations for which the geometry is shown in Figure 28. 
For each experiment, there were 25 minutes of suitable 
traffic data. 

7.2.2 Procedure 

General Characteristics 

The simulation has been run under varying conditions 
to test overall relationships. Figures 29, 30 and 31 
show macroscopic relationships between speed, flow and 
density in Lane 2 of a three-lane roadway. The 
generally known empirical relationships are satisfactorily 
reproduced. Figure 32 shows the capacity of the three- 
lane roadway as the percentage of trucks is varied, while 
Figure 33 shows the capacity of Lane 1 varying with the 
percentage of trucks in that lane. In both cases, the de- 
cline in capacity with increasing truck percentages is 
replicated. Figure 34 shows the capacity of an entrance 
ramp as a function of the length of the acceleration lane. 
The increase in capacity with increasing acceleration lane 
length is well replicated and indicates that the simulation 
is operating in a satisfactory manner for its high volume 
merges. 



155 




o 






o> 






H 
fa 

+J 

3 


o 




OU 


ID 




+> 


r» 




3 
O 

w 

> 

en 


o 


■ 


•H 


o 


4-> 


o 


vo 


IW 






^ 


CO 




tP 






c 


c 




•H 


<d 




o 


<D 




rt 


s 


o 


Qa 




in 


CO 


+J 


<=r 




3 

H 



o 
o 



o 



CN 

cu 
u 

tr 

■H 

fa 



Average Volume, Q 



156 





dj 




tn 




T3 




•H 


>1 ». 


c « 







u 


C 


PQ 







-P 


W r- 


01 


■H 





g 


u 


id — 


u a 




>H 




H z 


• • 


Q) >- 


vo i: 




c 


-p 


<d 


c 


X 


a) 


CD 


e - 


•H 


< 


M 




CD 


U 


& CD 


X 


> 


W 








X 


<D 7 


■P 


H 


<d 


^ 


-P 


"— ' 


CO 






>i 


C 


£ 


U M 


Q) 


CM 


a 




-P M 


P 











CO 


u 




XI 




£ 


• • 





CN <t 




rd 


-P 


CD 


c s 


<u 




e -p 


•H 


rd 


U 






& 


X ^ 


W 


04 




n 




X 




a 




H 




04 




^ 









m 




c 




<£> 


■H 




■P 


ia 


id 




^ 


CN 


3 




tr> to 


•H 


-P 


M-l 


C 


c 


CD 





e 


u 


•H 




M 


Q) 


0) 


C 


a 


<d 


X 


J u 


•# 




en 




(N 




CD 




H 




3 









•H 




&4 






157 



o 
o 



• • 



.•••:«. 



o 
o 
o 

CN 



CU 




E 




3 


• 


<-\ 


• 





« 


> 


• ••« 




• 


to 




> 


V 




• 


13 


• 


CD 




Q) 




a, 


• * . •• 


W 


• • 
• • • 




* • • 


s. 


• 


CN 


* • 


CD 




C 


• 


CO 


• 


1-3 


* 



o 
o 



O 
O 
CN 



O 
O 
00 



to 

•H 

x: 

(0 

c 
o 

•H 
+J 
ftj 

cH 

xi d) 

> 

CD 

- e 



CD 

s 



> 




> 
I 

CD 
CD 

CO 



CN 

d) 
U 

CT 
•H 



O 
O 



o 



o 



o 

CO 



o 

CN 



Speed , u , mph 



158 



>i 
-p 

•H 

CO 

c 

Q 



CO 

> 

<D 
<D 
O. 
CO 



a) 



o 



o 
in 



o 



o 



o 

(N 



— r- 

o 



o 







CO 






a 






•H 






x: 


o 




CO 


o 




c 


l-i 





•H 
■P 




e 


iH 




a 


a) 




> 


« 


o 


^ 


>i 


00 


^ 


•H 




^ 


CO 




>i 


C 




-p 


0) 




•H 


Q 




CO 


1 




c 


T3 


o 





a) 


V£> 


a 


• • 
3 


O 






T 




0) 

•H 
&4 


o 






CN 







Speed, u, mph 



159 



o 

CN 



>1 
■M 
•H 
CO 
C 
<D 
Q 

co 

> 

0) 



O 

> 



CN 

0) 

c 
h3 



o 
O 



o 
co 



o 



e 
> 



•P 
•H 
CO 

c 

(D 
Q 



o 



CO 

D. 

•H 
XI 
CO 
G 
O 
■H 
•P 
rO 

rH 

0) 

■P 
CO 

c 

' 0) 
Q 
I 

c; 

% 

H 



CO 



o 

CN 



at 
u 

ty 

•H 

Cm 













.... 












i . . 


1 

o 


O 


O 


O 


O 


o 


o 


O 


O 


O 


O 


o 


CN 


O 
CN 


rH 


CN 
iH 


00 


<tf 



Volume, Q, vph 



160 



00 



in 



CM 



(0 

0) 
<D 
U 
fa 

c 
o 

w 
u 

Eh 



CTv 



VO 



(0 
+J 
O 
E-< 



U 

»H 

Eh 

«4H 
O 



•4J 
C 
0) 
U 

>H 
d) 

w 
> 

0) 

E 
P 

rH 
O 
> 

(0 

0) 
0) 

Cm 



en 



CM 



o 
o 



«£> 



O 
O 
O 



o 

o 

in 



o 
o 

00 

in 



o 
o 
r» 

in 



o 
o 

in 



3 Lane Volume, Total (VPH] 



0) 

u 

p 

■H 
fa 



161 



00 




H 


^ 




C 

fd 

>1 
5 


o 




(U 


<tf 




n 

En 

G 


W 


CM 




o 


ro 


— • 






Q) 


En 




C 






(d 


+J 




J 


C 
0) 




* 


u 




(0 


M 


<s< 


^ 


CD 


CM 


3 


P4 




M 


(0 




En 


> 




M-l 


CD 











C*P 


i-i 


vo 







rH 




> 

Q) 
C 

•• 


00 




en 
ro 

CD 
H 

tn 

•H 



O 


o 


o 


O 


O 


O 


o 


«a< 


00 


CM 


VO 


O 


<r< 


00 


r~- 


r- 


\0 


ȣ> 


rH 


rH 


rH 


rH 


rH 


rH 



Volume, Lane 1 



162 



o 
o 



O 
O 



O 
O 



o 

o 

CI 



o 
o 



o 
o 



o 
o 
o 







4-1 









o 




C 


in 







r- 




•H 
-P 
U 




CD 


c 




C 


p 




fO 


fc -C 




hi 


4-> 
«0 Cn 


- 


c 


C 


o 





W CD 


o 


•H 


03 i-q 


vo 


4-> 






rd 


>i CD 




5h 


4J C 




<u 


•H f0 




■H 


O J 




<D 


(0 


- 


O 


a c 


o 


u 


fd o 


m 


< 


U -H 


■^ 




-M 




*+-l 


a td 











x: 


a; ^i 




+j 


<u 




Cn 


.. u 


" 


c 


^ u 


o 


<u 


co ^1 


o 


^ 




n 




CD 

3 
D 
•H 


^ 




Em 


o 






in 







Ramp Volume 



163 



These results are not intended to quantify the specific 
empirical relationships, but rather to validate the overall 
structure of the simulation by showing the known trends and 
relationships are present. 

Weaving Sections 

The simulation was run for the 25-minute period using 
the input lane flows and speeds from each of the two PINY 
empirical experiments. The outputs of the simulation were 
then checked against the field data. Generally, the per- 
formance of the simulation model was excellent. Volumes 
and total lane changes were reproduced with great accuracy 
while speeds and lane change positioning, although not 
exactly reproduced, were structurally as good as could be 
expected. Two particular characteristics were checked as 
significant indicators that the simulation is able to 
replicate given freeway weaving behavior. The first was 
the intensity and structure of the lane changing between 
individual lanes. It was essential, of course, that the 
numbers of lane changes be in good agreement with the 
empirical base, and this was obtained. In addition, the 
lane changes were distributed longitudinally through the 
weaving section in a reasonable approximation to the field 
results. Perfect statistical agreement was not an objec- 
tive since this would depend on local geometric and behav- 
ioral characteristics not specified in the data base. The 
second and most important weaving feature looked for was 
that the speed of the traffic stream be maintained through 
the weaving section at the given lane volumes and lane 
changing intensities. Any significant failing of the simu- 
lation would probably be first noticed in the forced lane 
changing process with consequent speed breakdown in high 
volume weaves. The fact that the simulation did compare 
reasonably well with the data for the output lane speeds 
(the input speeds being more closely controlled) , was the 
major validation factor in this phase of the study. 
Exact statistical agreement was again not expected since 
the precise speed patterns depend on factors such as 
vehicle type and desired speed which were not available 
in the base data set. 

The ability of the simulation to replicate these com- 
plicated, high volume weaving sections is significant, 
for it is this area of freeway performance that a simula- 



164 



tion has most difficulty accommodating. The results 
exceeded our expectations, and indicate the power and 
flexibility of the new simulation program. 

Each data set (Experiments 2 and 6 for the locations 
shown in Figure 28) was run for the 2 5-minute period. 
Input lane volumes and speeds were constrained to the data 
by the choice of generation parameter inputs, and then 
output lane volumes and speeds were compared between the 
simulation and the empirical results. The results are 
summarized in Tables 4 6 and 4 7 and they indicate that very 
good agreement was reached. 

To test the overall agreement of the empirical and 
simulated lane specific values of the various traffic 
characteristics, a correlation analysis was performed. If 
the agreements were exact, then a correlation coefficient 
of +1.00 would be obtained. Table 48 shows the summary. 
Excellent agreement was obtained for all comparisons. 
As explained earlier, exact statistical fits were not 
appropriate since many key driver and vehicle parameters 
were not known. The correlation coefficients of .74 and 
.68 for speed outputs, for example, are considered to be 
excellent in the light of the potential for variation of 
this particular characteristic. 

For Experiment 2, a series of runs statistically 
checked the operation of the simulation against the 
empirical base data using the standard t-test. These 
characteristics were tested statistically since the para- 
meters were closely associated with the simulation's 
vehicle generation process. 

Three separate tests were conducted: 

(1) Lane inputs, 

(2) Speed inputs, 

(3) Lane change totals. 

The data used for the tests were obtained from a steady 
input model of Experiment 2, where the inputs for the first 
5-minute interval were used to generate ten 5-minute 
intervals. Thus, the sample consisted of ten generated 
values and the t-test determined whether the mean of these 
values was significantly different from the empirical data. 



165 







en 






















(0 




4J 






















-p 
id 




3 

a. 


1C 

■u 


CN 


"3- 


'S' io 


cn id 


cn en 


(M r- H <J 


vo © co r» 


r-i in © rH 






4-> <0 


co 


■» 


co co 


CN CO 


(N CN 


co oo oo co 


CO ^" CO CO 


CO CO CO CO 




Q 




3 Q 
























o 






















T) 




"0 


H 

3 p 
°-H 




















■H 




<D 


oo r- 


id cn 


r» co 


oo r- 


© co io in 


00 CO 00 cn 






0) 




CO 


cn co 


co cm 


cn oo 


CO (N 


oo co co cn 


CO CO CO cn 


r- ^ ~~o oo 






a 
















CN CO CO CN 




-H 




w 


Lj 




















Eh 




< 




















73 




01 
4J 


a 

3 Q 














* 


* 

CO © t~ CO 




rtJ 




3 


CO 


in 


o r» 


oo r» 


ID CO 


o id oo in 


O O rH CJ> 


CN CN CO 






a 
c 


co 


n- 


co oo 


co oo 


cn cm 


vfNn 


^* co co 










M 


s 




















CO 


























-p 

3 




-a 
a) 


6 


r» 


CN 


© i-l 


in oo 


O i-l 


© TT O CO 


in co cn .H 


r-aiOi-t 


CO 




<o 


-H 


co 


■W 


co co 


oo <» 


co co 


«W "3* 00 01 


CO T CN CO 


CO rH CO CO 


3 


a, 
-p 




M 


CO 


















rH 

(8 

> 


3 
























o 




»1 
4J 
3 


(0 




















u 


G 




a 


+J 


r> 


O 


r- o 


'r in m in 


CO rH CO CO 


O © CTi >J3 


ID CN •q- vo 


N 







-P 


Id 


r- r~ co r» 


CO en 00 CO 


00 VO 00 00 


in r^ id r>- 


00 rH ID 00 






3 


— Q 


















<a 


•H 




O 


CO 
4J 




















-P 






















CQ 

s 

o 


<d 




£ 
a 


c 


TT 


CTi 


in er> 


in tn 


00 © 


co cn r- o 


r-l o co in 


[-* ^ co in 


D 




rH 


r~ 


in 


in id 


cn en 


in co 


00 ID ID 00 


VO CN r~ ID 


CO rH ID ^< 


CO 


e 



























> 




















(0 


•r-l 


V£> 


"^ ID 
+J 

.. 3 


















■p 


CO 




















10 


M-l 


-P 


(0 

3 


•H * 

e (0 


** 


cc 


in cn 


Cn cn 


rH rH 


CN CO in rH 


* 
in © cn in 


* 

© O ID © 


Q 





<D 


a, 


r~ 


KD 


(n in 


r» cn 


CN <D 


oo in i-i in 


CO O «3" 


O O ID 


C8 


e 


c 

H 


£ Q 






rH 




r-l 


rH 


r-l 


rH 


u 

•H 


C 


•H 






















14 



CO 




e 

3 


6 


in 


u-i 


[-. tt 


in \d 


<Tl CN 


(MH iOO\ 


CO rH CO CO 


CO rH O •fl- 


■H 

a 
6 


•H 


a 


.-i 


■H 


r- 


LI"! 


O "S- 


r-» © 


rH T 


oo in © co 


P- rH CTl CO 


cn rH ^r 


61 


u 

at 


X 




> 


w 






rH 


rH 


rH 


rH 




rH 


* 



e n 
o o 

U Mh 






cu 

■H 
XI 

(d 



10 


n> 




* 






<o 






(0 




> 


> 




> 






> 






> 




1-1 


u 




u 






m 






^ 




a 


a 




CO 






a> 






CO 




jj 


4J 




4J 






+j 






V 




c 


c 




c 






c 






c 




•H 


•H 




■H 






•H 






•H 




. rH CN CO 00 


• rH 


CN CO CO 


. 


pH 


CN CO CO 


. 


•H 


CN CO 00 


• rH 


CN CO CO 


C 


C 




c 






c 






c 




•h cu cu co a 


■H CO 


co co a 


•H 


0) 


oj co a 


•H 


<D 


co co a, 


•H CO 


co co a 


e c c c g 


g c 


C C g 


g 


c 


c c g 


g 


C 


c c g 


g c 


c c g 


1 (t) fl <0 ifl 


1 * 


10 IT) (Q 


1 


id 


(a cj n 


| 


« 


io io id 


1 <a 


<0 (0 <0 


in J JJ2 


m .J 


J h1 K 


m 


J 


jja 


LO 


J 


JJK 


in j 


J Jo: 


+J 


T3 




-O 






JC 






JS 




W 


c 




V-l 






■u 






4J 








166 



u 
o 

<4-l 

(TJ 
■P 

rfl 
Q 

T3 
iH 
0) 
•H 
fa 

c 

fO 

en 
-P 

a, 
-P 

o 

o 
•p 

■H 

6 
•H 
CO 

O VD 

C -P 

O C 

co 0) 

■H g 

a cu 

e a 

o x 

u w 



t-- 

■** 

0) 

rH 
XI 
fO 

Eh 



01 










4J 










3 


<o 








a 


4J 


T 


r^ 


iH in 


4J 


(0 


■<* 


*r 


in ^> 


3 


Q 








O 


,_^ 










u 

li 


rH 


r» 


t in 


a) 


'a' 


^r 


•q> PO 


w 


M 
















in 


a 








4-1 


«5 

rH w 








3 


in 


oo 


i-i m 


a 


t 


«* 


in ro 


c 


3 a 








H 


s 








"0 


• 








0J 


a 


"d" 


r- 


<-t © 


0J 


•H 


t 


T 


TJ- TJ- 


a 


co 








w 











cfl 

4J 

^ "3 

W Q 

4J 

c 

3 | 

4-1 

3 

c 

p IB 
65 4-> 

So 



CO., 



CO CN CO 00 

r- t r~ cn 



tt po m r- 
vx> <n in r» 



rr rH cn co 
t vo [» H 



tj- rH in 03 

H in ID rH 



Hen n<s 

^» ^» ^" ^ 



f> (N oo ro 
ro T «3" ro 



^* ^ ^ ro 



co po r- ,h 

ro ^" ^ ^ 



in in ffi ^ 

Hinioo 



in m rH r- 
o in r- oi 



vnoo 
r~ kd r- po 



ro Ol rH o 

oo r~ r- cn 



CPi ro va (N 
ro ^* ^« ^« 



f-~ TJ> o O 

<N •q- in in 



HCllOO 
^ ^" ^" ro 



rH P- CN rH 

t t in t 



r- cn p» oo 

OflNO 
rH rH 



en in o id 
oo ro in ro 



rH r~ o o 
r~ in rH cn 



in (N (N lO 
00 ID rH CN 



r- OffifM 
ro ^ ro "3* 



cn in o o 
ro t in ro 



O O O ro 



CN r- rH o 



er. t t in 
o ro cm 



"3" cn ro rr 
ID rH cn ro 



rH r» O ID 
r- in .H 



P- <T H fl 1 

m *r rH 



O ro o in 
ro 



ro o ro ro 

IN TT TT rH 









Qi 








3 


o o 


o 


PO 


iH 


rr *r 




ro 


CO 
> 


u 


rH VO 


o 


CO 


0) 


«*• rr 




m 


N 

CO 
01 

3 




<J> rH 


o 


o 


cn ro 


rH 


H 


co 


rH 




rH 


CO 
4-1 
CO 
Q 


r- o 


CO 


T 


rH 


in po 


PO 


CD 


CO 


rH 







■H 

U 
•H 

a 


in pm 


o 


t"» 


E 


r~ in 




(N 


W 



m r« o u3 
id r- pm 



cO 






CO 










CO 








CO 








CO 










> 






> 










> 








> 








> 










u 






1-1 










M 








u 








U 










<a 






<u 










3) 








0) 








<l) 










4J 






■p 










■P 








4J 








4J 










c 






c 










C 








c 








C 










•H 






•H 










•H 








•H 








■H 










• iH 


CN PO 


CO 


. 


iH 


CM 


PO 


CO 


. 


rH 


CN PO 


CO 


. 


rH 


CN PO 


CO 


. 


iH 


CN 


PO 


CO 


e 






c 










c 








c 








C 










•H CJJ 


<D <U 


a 


•H 


0) 


<u 


0) 


a 


•H 


<D 


<U d) 


a 


•H 


(1) 


0) 0) 


a 


•H 


0) 


(1) 


gj 


a 


s c 


C C 


£ 


6 


c 


c 


c 


s 


£ 


c 


C C 


a 


£ 


C 


c c 


£ 


£ 


c 


c 


c 


£ 


1 ca 


CO 16 


CO 


l 


t3 


BJ 


CO 


cfl 


1 


CO 


CO CO 


CO 


l 


<0 


CO CO 


cfl 


1 


CO 


CO 


03 


CO 


m iJ 


>A »J 


OS 


in 


rJ 


J 


tJ 


« 


in 


J 


J tJ 


QJ 


in 


rJ 


J >J 


£ 


in 


J 


J 


J 


C£ 


4J 






T3 










-O 








J= 








.c 










UJ 






C 










M 








4J 








■W 










rH 






CN 










CO 








"3- 








m 











167 



Table 48: Correlation Analysis of PINY 
Data and Simulation Outputs 



PINY Experiment 2 



Lane Inputs 


r 


. 


.99 


Lane Outputs 


r 


= 


.96 


Lane Changes 


r 


= 


.98 


Speed Inputs 


r 


= 


.71 


Speed Outputs 


r 


= 


.74 


PINY Experiment 


6 




Lane Inputs 


r 


= 


.99 


Lane Outputs 


r 


= 


.94 


Lane Changes 


r 


= 


.98 


Speed Inputs 


r 


= 


.81 


Speed Outputs 


r 


= 


.68 



168 



Table 4 9 summarizes the results and shows that good 
agreement was obtained. 

Finally, Figures 35 to 38 illustrate the weaving 
intensities longitudinally through the sections. Each 
base data set gave lane changes by quadrant and accordingly 
these were checked against simulation outputs. The 
cumulative numbers of lane changes by quadrant are compared 
and good results are obtained. The exact position of the 
lane changes will depend on local geometries, signing, and 
traffic characteristics which cannot be introduced into 
the simulation. As a consequence, exact replication was 
not expected. The general performance of the simulation 
exceeded expectations. 

Headway Distributions 

Figure 39 shows headway distributions recorded in the 
simulation for free flow and congested flow. The simula- 
tion was run initially at free flow at about 1650 vehicles 
per hour per lane. The free flow headway distributions 
were recorded. Then, a bottleneck was introduced, and 
congestion set in with traffic flow now about 1500 vehicles 
per hour per lane. The congested headway distribution was 
recorded. The two distributions show clearly the charac- 
teristic differences with the congested distribution more 
normal in shape and shifted to the right. 

Figure 40 shows headway distributions for medium 
traffic flow reproduced by the simulation and compared to 
field data derived from the Long Island Expressway data 
set. 

The simulation gave a good representation of a well 
behaved headway distribution. The empirical data was 
less well behaved but reasonable agreement was obtained 
with a chi-squared test not significant at the 5 percent 
level. 

As indicated earlier, the Los Angeles closely spaced 
data set consisted of six 5-minute counts for each of 
16 detectors arranged in groups of four at 600 feet 
spacings. Thus, the total data set consisted of 72 5-minute 
counts with headways grouped by half- second interval. 



169 







CD 






-P 






rd 






U 






CD 


rd 

rd 




CD 


Q 







C 




-P 


a; 

> 

•H 

o 




T3 
CD 
CO 

P 


•H 
5 




(U 
-h 
CD 


c 



•H 

+J 

rd 

H 

e 

•H 




H 
rd 
> 

CD 
■P 

C 
•H 


CO 




<D 


4-1 





-P 

P 
C 


c 


(0 

•H 

rd 

a 
e 



u 


CM 


•H 

e 

1 CO 




ID H 


-P 


rd 


c 


-P > 


CD 


CO U 


e 


U CD 


•H 


•H -P 


CD 


iw g 

•H 




a 


<u 




X 


£ cd 


<J\ 


w 


■P -P 
P 


^ 




.H C 


0) 

rH 

CO 

Eh 


•• 


-H 


c 


•H 


Mh E 
1 

co in 


-p 

rd 


•P 

P 




u 


a 







c 




hH 


H 



•H 


>l 


• 






























4n 


H 4-1 






CO 










CO 












CO 


•H 


-p 4-1 








CD 











O 


CD 





o 


o 








CD 


C 


C -H 


S 


s 


>H 


2 


S 


s 


Z 


>H 


z 


S 


s 


z 


2 


>H 


C" 


ro a 






























•H 


u 






























CO 


































CO 


• 
































CD 


> 


<* 


r- 


H 


CO 




CTi 


oo 


<tf 


iH 


in 


o 


00 


VD 


in 




P 


CD 


vo 


in 


00 


^ 


H 


H 


oo 


"<* 


CN 


«3< 


r~ 


(Ti 


CN 


00 




H 


a 
































rd 


• 


H 


CN 


CN 


H 


rH 


00 


00 


CN 


on 


CN 


CN 


00 


CN 


H 




> 


T3 
■P 










H 










H 












TS 


CO 
































Q) 


































■P 


















^ 
















rd 


























o 


o 


o 




(-1 


C 


r^ 


CN 


O 


00 


H 


o 


H 


CN 


tH 


o 


CN 


V£> 


oo 


<T\ 




CD 


rd 
































C 


CD 


■^ 


oo 


in 


^ 


<£> 


H 


on 


«* 


c^ 


>x> 


CTi 


^* 


in 


m 




CD 


g 


«* 


"* 


»* 


00 


"* 


«^) 


00 


H 




<T\ 


H 










O 












H 


tH 


tH 

















C 


CD 


00 OO 00 rf 






CD 


p 


• • • • 


^ H <Ti CO 


CO 1/1 Cf\ -* 


> 


tH 


in oo o in 


•"3" V£> t> H 


CO H 


•H 


ftj 


<* ^ m oo 


tH H H 




u 


> 









in o 





T3 


'a 


tJ 


•a 
























CD 


CD 


CD 


CD 
























CD 


CD 


CD 


CD 










H 


00 


H 


CN 


CN 


00 




a 


a 


Cm 


Cm 










1 


1 


1 


1 


1 


1 


M 


co 


co 


CO 


CO 










00 


H 


CN 


H 


oo 


CN 


CD 










CD 


CD 


2 


CD 














-P 


■p 


-p 


-p 


+J 


e 


S 


6 


e 


CO 


CO 


CO 


CO 


CO 


CO 


0) 


p 


p 


p 


p 


p 


P 


P 


p 


CD 


CD 


CD 


CD 


cu 


CD 


e 


a 


a 


a 


Cm 


H 


iH 


iH 


H 


cn 


Cn 


Cn 


Cn 


Cn 


Cn 


rd 


c 


c 


c 


c 














c 


C 


C 


C 


C 


C 


u 


H 


H 


H 


H 


> 


> 


> 


> 


fd 


rd 


rd 


rd 


rd 


rd 


rd 


















Xi 


JC 


£1 


X! 


£ 


X 


Cm 


tH 


CN 


00 


CO 


H 


CN 


oo 


00 


U 


U 


U 


U 


u 


U 




CD 


CD 


CD 


& 


QJ 


CD 


CD 


& 


<D 


CD 


<D 


CD 


cu 


CD 




c 


C 


C 


B 


C 


C 


C 


6 


C 


C 


C 


C 


C 


C 




rd 


fd 


rd 


rd 


fd 


cd 


rd 


rd 


rd 


rd 


rd 


rd 


rd 


rd 




hH 


hH 


J 


tf 


J 


HH 


KM 


Cm 


HH 


hH 


hH 


hH 


hH 


hH 



170 



\ \ 






\ \ 






\ \ 




CO 


\ \ 




0) 


\\ 




> 




\ 


.h 

Q) 
•H 




\ 


fa 


? \ 






H \ 

<3 1 




\ 


> ' 






S-l 




\ 


Q) 






•P 




\ 


Ch 






e 






\ 


u 




V \ 

\ \ 
\ \ 

\ \ 
\ \ 
\ \ 



ro 



CN 



<N 

4-> 
C 
0) 

e 

•H 

U 

x 

w 



i 

CM 

W 
Q) 

C 

I 

w 
a) 
en 
C 

fd U 
3 



-P 
c 



a 



in 

CM 



o 

CN 



in 



in 



0) 

c 
<d 

4-1 
O 

c 
o 
w 

•H 

(d 

e 
o 
u 



in 
m 

0) 
U 

tP 
•H 
fa 



Number of Lane Changes (5 min) 



171 



CN 



\\ 








\ \ 








\\ 


V 








\\ 


(0 

3 






\ \ 


r-i 






«0 






\ \ 


> 
u 






\ \ 


Q) 






4J 






\ \ 


3 






\ 


\ o 
\ u 






\ 








\ 


\ x 






w 


\ \ 









\ \ 






■H 


\\ 






> 








TJ 








.H 








<D 




N^ 




•H 




NN 




fa 







ro 



CN 



c 
a 



+J 

c 
<D 

e 

•H 

a 
w 

I 

(0 

-p 

c 
a) 

e 

CD 
> 
O 

g 



(0 

8. 

C 

«a 
jc 
u 

<u 

o 

c 
o 

10 
•H 

6 
o 
u 



WD 

n 

<D 
H 

Cn 
•H 
fa 



o 



o 

CN 



O 



o 



o 

m 



Number of Lane Changes (5 Min) 



172 



X 








V 










\\ 




CO 




\\ 




CU 

3 




\\ 




.H 




io\ 


\ 


> 




<D \ 




T5 




3 \ 

rH > 




\ 1 




> 




\S 




u 

CU 




\ \ 




4J 




\ \ 




£ 





\ \ 




u 




\ \ 
\ \ 
\ \ 

\ \ 

\ \ 

\ \ 



c, 



CM 



-p 

C 
(0 

<d 
a 



o 

Lf) 



O 



o 



o 

CN 



4-> 
C 

<u 

e 

-H 
CU 

a 
x 
w 



00 
I 



CO 

CU 

c 

i 

to 
cu 

(0 
£1 
U 

CU 

c 

o 

c 
o 

CO 

•H 

e 
o 
u 



m 

CU 

en 

•H 

fa 



Number of Lane Changes (5 Min) 



173 



VD 




«tf 



ro 



-P 

C 
rd 
U 

fO 

P 

a 













o 


O 


O 


o 


O 


IT) 


O 


in 


o 


in 


CM 


CM 


H 


iH 





•p 
c 
<u 

e 

•H 
CD 

a 
w 



(0 
4J 
C 

'D 

e 

CD 
> 


a 



< 

i 

CO 

0) 

en 

C 

U 

CD 
C 
CO 
iJ 



C 

o 

CO 
•H 
H 
rd 

a 
£ 
o 
u 



oo 
ro 

CD 

CP 

•H 

fa 



Number of Lane Changes (5 Min) 



174 



o 



O 

o 



<^ 






CO 




C 









-H 




4J 




3 S 




X) 


• 


•H rH 


u 


>H (i, 


<D 


-P 


CO 


CO T3 




•H <D 


CO 


Q +J 


>i 


CO 


rfl 


>i CD 


5 


(ti Cn 


■a 


5 c 


fO 


T! 


cu 


fO U 


a 


cy 




k >d 


CD 


c 


e 


fO 


•H 




Eh 


<* Si 




m cu 




h 




CD l*i 




^ 




3 




Cn 




•H 




fa 



Frequency 



175 




"55 S" 

• • . 

o o o 

Aq-TiTqeqojct 3ATq.exnuino 



o 



176 



The simulation was run for a 30 -minute period with the 
same detector spacings. The generated input flows used 
were the overall lane averages determined from the empiri- 
cal data. No attempt was made to generate the detailed 
5-minute variations since exact validation at this fine 
level would also require knowledge of driver types, ve- 
hicle types, and geometries, including ramp locations. 

A chi-squared test on the 72 5-minute counts indicated 
that 14 of them were significant at the one percent level. 

This is considered to be a most satisfactory check in 
view of the fact that relevant information is lacking from 
the empirical data and the simulation was run at a general 
level without fine tuning to the detailed empirical 
variations. 

Figures 41 to 44 show the comparison of the simulated 
and empirical headway distributions over the full 30-minute 
period for each lane at one detector set. The agreement 
in the structure of the distributions is very good. The 
agreement in the crucial area of short headways is also 
close. 

Trajectories 

The major validation here was the simulation of a 2 3- 
vehicle platoon through the shock wave indicated by the 
Ohio State data. The leader of the platoon was given the 
exact speed and position throughout the time period of 
50 seconds as given by the empirical data. The 22 fol- 
lowers were started with the speeds and positions as given 
by the data set. The behavior of the platoon was then 
simulated for 50 seconds. The driver types were adjusted 
to develop the correct individual following characteris- 
tics, while an average vehicle type was assumed. 

Figure 4 5 shows the result. The lead vehicle and 
last vehicle in the platoon are diagrammed along with 
every fifth vehicle. Agreement between the simulation and 
the data is very good. The main variation is for the 
later vehicles as they enter the shock wave where the simu- 
lated vehicles tend to decelerate a bit earlier. It 
should be remembered that any variations by an individual 
vehicle will be propagated through all followers. In 



177 




<N 



- 00 



- KO 



- <tf 





o 




c 




<d 




^ 




M 


<-^ 


CO <1) 


CO 


>iT3 


T3 


rd rH 


C 


S 3 





'O 


O 


(d x: 


CD 


CD CO 


CO 


IS 


*— ' 


VD 




W LO 


CQ 


CD =*= 


>i 


.H 


fd 


cd u 


S 


C7> 


T3 


C 4-» 


(d 


< u 


(1) 


Q) 


ffi 


CO +J 




CD 




J Q 



_ CM 



CD 
U 

CTi 
•H 



Cumulative Proportion 



178 



<a 




4-> 




<0 




Q 






c 


rH 





cti 


•H 


U 


-P 


•H 


«j 


M 


rH 


•H 


3 


CL 


e 


e 


•H 


w 


CO 




CN 



CO 



. ^£5 



- ""a" 



- CM 





<u 




c 




CO 




k! 




M 




(U 




'O 




iH 




D 









x; 




w 




CO 




>l-U 


^— . 


fO 


CO 


5 -P 


T3 


T3 X 


C 


fd cu 


O 


<D S 


O 


EC 


cu 


in 


CO 


en m 


*-* 


CD =** 




iH 


w 


0) to 


>i 


tn o 


(0 


C -M 


? 


< o 


T3 


(D 


tf 


W -P 


CD 


O CD 


a: 


^ Q 




• • ' 




<N 




^ 




a; 




u 




3 




tn 




•H 




fe 



1 


1 — 


1 


» 


f 


O 


00 


VD 


T 


CN 


• 


» 


• 


• 


• 


iH 


o 


o 


o 


o 



Cumulative Proportion 



179 




CO 

CJ 
O 

u 
<d 

CO 



CO 
CD 

a 



CD 

c 

hq 

c 
fd 

•H 

'd 

CD 

s 

en o 

£ -p 

T3 X 

fd 0) 
CD 2 

co LD 
CD =4*= 
rH 

CD U 
Di O 

C -P 

< o 

CD 
CO +J 
O CD 
J Q 



^' 

CD 
U 

•H 

fa 



r 

00 






CN 


• 


• 


• 


• 


o 


O 


o 


O 



Cumulative Proportion 



180 




CM 



_ 00 



. \£> 



- ■* 





CD 




e 




<d 




J 




co 




>i c 




tO (0 




S -H 




T> »Q 


. — . 


rd Q) 


10 


0) S 


>d 


K 


c 


CO 





co in 


u 


CD =M= 


CD 


H 


w 


CD >-l 


* — 


tJi 




C -P 


CO 


<C O 


>1 


CD 


fO 


CO 4-> 


25 


<D 


T3 


J Q 


(0 




CD 


• • 


S3 


•* 




si 1 




CD 




U 




P 




fci 




•H 




En 



- CN 



Cumulative Probability 



181 



8000" 



7000- 



+J 

Q) 

Q) 



6000- 



5000- 




Field Data 
Simulation 



-I 1 r 

20 30 40 

Time (Seconds) 



— r 

50 



10 



Figure 45: Ohio State Vehicle Trajectories Platoon 
of Twenty Three Vehicles Showing Paths of Ve- 
hicles Numbers 1, 5, 10 , 15, 20, and 23. 



182 



examining the trajectories in detail, it was noticed that 
even for this comparatively short time period, individual 
drivers tended to change their "type" of car-following 
behavior; i.e. , their desired following distance. 

Nevertheless, despite these latter factors and a 
lack of knowledge as to the local geometries that create 
this bottleneck, this simulation agreed most satisfac- 
torily with the empirical base. 

Other platoons were run for the free flow trajectory 
sections and in these cases the simulation-empirical match 
was so close as to be barely dif ferentiable at a level of 
detail as given in Figure 45. 

8. SUMMAR Y 

The program design described in the preceding sections 
is implemented in a CDC 7600 computer at the Brookhaven 
National Laboratory, Upton, New York. This implementation 
includes the adaption of the component models, described in 
Sections 5 through 7, which were originally programmed 
for a DEC computer system at the University of Pittsburgh. 
The simulation model is also operational on an IBM 360 
computer system, similar to that at the FHWA computer 
facility in Washington. This development procedure ensures 
that the INTRAS model may be exercised on a variety of 
computer systems, as has its predecessor, UTCS-1. 

Validation of the INTRAS simulation model (as reported 
in Volume III) is accomplished by comparison of simulated 
detector outputs with the closely spaced detector data 
base from the Los Angeles freeway system. The high 
density of detector stations (600-foot separation) , in 
this data base, makes it possible to rigorously test the 
INTRAS model's ability to replicate abrupt longitudinal 
changes in traffic performance, characteristic of congested 
flow conditions. Comparisons are also performed of model 
performance with on-ramp field data gathered in the 
Washington, D.C. vicinity. 

In addition, the PINY data sets used for component 
validation (Section 7) are reapplied to testing of the 



183 



integrated model to ensure that component performance is 
not affected by the integration process. 

The results of validation testing, as well as 
exercises performed to evaluate and demonstrate the INTRAS 
incident detection capabilities, are described in Volume III. 



184 



APPENDIX A 



185 



INTRAS 
Supervisor 



Case 



Over 



Initialize Flags 




Last case has 
been processed 



s~\ Rew 



Read and Test 
Run Control Card 




rror Count > 



Determine Next Module 
To Be Called (NEXCAL) 




Call Module 


Specified 


by (NEXCAL) 




NEXCAL 


MODULE 


1 


PORGIS 


2 


LIS 


3 


SIFT 


4 


FUEL 


5 


INCES 


6 


POSPRO 


7 


INPLOT 


8 


SAM 



I 



Figure 4 6: INTRAS Supervisor Logic 



Rewind All 
.es 



I 



Call FIN if 
CALCOMP Tape 
has been 
created 



END 



186 



PORGIS 



Table 



Contents Request 



Zero Error Count 
Set Last Subinterval 
Flag, LASUBF=0 



Read Case Data and put on 
utility File 13. Fill 
ICDCT with card counts 
(Types 00 thru 03) 
Print card list 



Create 
Table of 
Contents 
for Case 
Data Tape 



Set Case 
Over Flag 
NEXCAL=9 



T 



Return 



Rewind 13 



Call CLRALL 
To clear or reset 
appropriate 
arrays and scalars 



©■ 



IRSCT=IRSCT + 1 



IRSCT > 1 



Read Title Card 
and Check 



Read Network Name 
Card and Check 



Call LINKIN 
To read and test individually 
card types 2 , 3 , 4 , 5 , 6 and 7 
The appropriate arrays and 
scalars are primed. The link 
array sizes are revised 
appropriately . 



6 



-»*T 



Figure 47: PORGIS Module Logic 



JC7 



Call INT1 
To prime system 
geometric arrays 
and scalars and 
perform system 
testing 



Call PRSIG 
To prime signal 
and sign control 
and test 
diagnostically 



IRSCT=1 



Call TURNIN 
To read and check 
Link turning movements 



IRSCT=1 



Call 


PRACT 


To prime 


traffic 


response 


inter- 


section 


control 


and test 


diaa- 


nostically 




Call PRMSND 
To prime input 
flow rates and test 
diagnostically 




RSCT > 1 



Call SURVIN 
To input 
surveillance 
detectors and test 
diagnostically 



Call INCIN 
To input 

specifications of 
incidents and test 
diagnostically 



Figure 47 (continued) 



188 



Call IMBED to input revisions 
to imbedded calibration data 
and test diagnostically 




IRSCT > 1 



Read Card Types 50, 51, 55; 56 and" 57 
(specification of plot, incident out- 
put and on-line incident processing 
data) and test diagnostically 



Call DETGEN to generate 
detectors for contour 
plot output 



Read simulation control 
card (type 60) and 
test diagnostically 



ZE 



IRSCT > 1 



Call LINOUT, SIGOUT, FLOOUT, 
SUROUT, INCOUT, and IMBEDO 
to print the input tables 



Print User Options 




Last Subinterval 



Store Data 
on Case 
Data Tape 



Set last subinterval 
flag, LASUBF = 1 
IRSCT = 



o 



:erct=o 




Data not to 



be stored on 



tape 



ERCT>0 



RETURN 



Figure 47 (concluded) 



189 




PORGIS has already stored 



card data on File 13 



Retrieve data from Case Data 
Tape and store on File 13. 
Fill ICDCT array with card 
type counts 



Set last subinterval 
flag, LASUBF=0 



Rewind 13 



IRSCT=IRSCT+1 




IRSCT > 1 



Call LCLRAL to clear or reset 
appropriate arrays and scalars 



Read Title Card 



Read Network Name Card 



Call LLINKI to read card 
types 2,3,4,5,6 and 7 . 
The appropriate arrays 
and scalars are primed. 
The link array sizes 
are revised appropriately 







Figure 48: LIS Module Logic 



190 



Call LINT1 to prime 
system geometric 
arrays and scalars 


„ IRSCT= 


=1 ■ > 


> 












v 






Call LTURNI to read 
turning movements 



Call LPRSIG to prime 
signal and sign 
control 



IRSCT=1 



Call LPRACT to prime 
traffic responsive 
intersection con- 
trol 




Call LPRMSN to prime 
input flow rates 




IRSCT > 1 



Call LSURVI to input 
surveillance detectors 



Call LINCIN to input 
specifications of 
incidents 



Figure 48: LIS Module Logic (continued) 



191 



i 



Call LIMBED to input 
revisions to imbedded 
calibration data 




IRSCT > 1 



IRSCT = 1 



Read Card Types 50, 51, 55, 
56 and 57 (Specification of 
plot, incident output and 
on-line incident processing 
data) 



Read simulation control 
card (Type 60) 



PQRGIS has already 



printed input tables 




IRSCT > 1 



Data originated 
from case data tape 



Call LLINOU, 
LFLOOU, LIMBDO 
to print 
revised input 



Call LLINOU, LSIGOU, LFLOOU, 
LSUROU, LINCOU, and LIMBDO 
to print the input tables 



Print subinterval 
length and current 
user options 




Last 



Subinterval 



Set last subinterval 
flag, LASUBF=1 



RETURN 



Figure 48: LIS Module Logic (concluded) 



192 



SIFT 
Supervisor 



Call LOCON 
ISSEO=l 



Start Control 
loop 



Call HICON 



routing 




Time to perform 



incident detection procedures 



Call LOCON 

ISSEO=0 
i 



Time to Re-Evaluate 



Call LOCON 
ISSE0=1 



Contour Time 



step over 



Call LOCON 
ISSEQ=5 



Standard 



O Report 
Time Step 



Call LOCON 
ISSE0=6 



Over 






Detailed 
Report 



Time Step 
Over 



Call LOCON 
ISSE0=7 



Fill 


Call 


LOCON 


Evaluation 


ISSE0=2 


Cycle Over 


'' 



ISRET=2 < ^\ > 

Equilibrium SjZ 



Equ 

Not Attained 



Next Time Step 



G 



Figure 49: SIFT Supervisor Logic 



ISRET 
= 1 



193 



© 



1 


\^ill Time 
[^Completed 


Jusi^ 


* ^Simulate (Ec 


milifc 


rium 




■ 


*r Attained or 
Irrelevant) 




Determine 
Next 
Module 
Call 


Case^x^ 


Do not 
Simulate 


' 


r 




^"S 




Call LOCON 
ISSEQ=3 


Over^ 


y 

Subinterval 
Over 












< 






1 




Set Case 
Over Flags 


© 






i 


' 




1 










Next Module is 
LIS or POSPRO 






< 


' 








' 






\ 


' 










RETURN 







Figure 4.9:SIFT Supervisor Logic (continued) 



194 



( LOCON J 



ISRET=ISSEQ 



Branch on value^-*vof ISSEQ 



_L 



Call SINCES 
to perform 
incident 
detection 
procedures 



Call FILTST 
to test for 
inflow/outflow 
equilibrium 




1'hrium 



ieved 



Call INIT 
to perform 
initialization 



K? 



Call RESET to 
prepare for 
simulation 



<> 



Start 
of fill 
time 



Call lrank to 
order freeway | 
links 



Call CYCP 
to print 
standard 
report 



Call CPTOUT 
to output 
MOE values 
for INPLOT 

contours 



Call FTSC to 
determine 
freeway time 
step 



[Set vehicle 
processing 
I counters 



|I3RET=1 | 



Call INTST 
to print net 
work status 
reDort 



I RETURN 



J 



Figure 50 : LOCON Suboverlay Logic 



195 



HICON 



Call MOEV to move existing 
non- freeway vehicles 



Call SVEH to generate new 
non- freeway vehicles on 
entry links 




Freeway Time-Step 



not over 



Call FMAIN to simulate 
freeway for one time-step 



Update freeway clock 



Call UPSIG to update 
signal control 



Call CLNUP to adjust statistics 
and perform other bookkeeping 
functions 




Vehicle 



rajectory 
time-step 
over 



Call TPTOUT to output 
vehicle trajectories 
to INPLOT Data tape 

f 



RETURN 



Figure 51: HICON Suboverlay Logic 



196 



POSPRO 




First Subinterval 



just concluded 



Establish 
temporary 
statistical 
File 



Add current 
subinterval 
statistics 
to temporary 
file 



More 



Final 

subinterval 

concluded 



subintervals 



Next Module 
is LIS 



Copy Temporary 
file to Statistical 
data tape 



Determine next 
Module call 



RETURN 



Figure 52 : POSPRO Module Logic 



197 



INPLOT 



Read INPLOT Request 
Card Type 70 



o 



index 



e quest 



Call FILEX to 
print index of 
cases on plot 
data tape 



vo 



a 



contour 
„ value 



change 
request 



Read in and 
store revised 
contour values 



last 


Determine 


request 


next 
module 




v 




RETURN • 



Call IOPROC 
to read data from 
plot data tape 
onto disk files 



contour 



request/*\t 

*— :z <s> 



rajectory request 



check trajectory 
data time limits 



Check MOE data 
time limits 



All ICE's 




Selected 



MOE 



loop over 
all MOE's 



Call 

CONTR 



All lane 



Call CONTR to 
generate plot 
for one MOE 



o 



selected lane 



Call SPTAL 
to generate 
trajectory plots 
for all lanes 



Call TRAJEC 
to generate 
trajectory plot 
for one lane 



Figure 53: INPLOT Module Logic 



198 



INCES 



Read INCES Parameter Cards 
Types 65, 66, 67 and 68 



I 



Position INCES Data tape 
at beginning of file 



Point Processing 



Request 



r < x^ Ss x ^ MOE Estimation Request 



Point Processing 



Incident 

Detection 

Request 



required and not 

previously 

performed 



Call POINT to 
generate 
Detector- 
specific 
evaluations 



Incident 



Detection 
Request 



Call appropriate subroutine 
M0E1, M0E2 or M0E3 to per- 
form MOE estimation procedure 



Call appropriate 
Subroutine INC1, 
INC2 or INC3 to 
perform specified 
incident detection 
procedure 



Call POUT to 
print point 
processing 
report 
1 



Call MOUT to print 
MOE estimation 
report 



Call PINC to print 
incident detection 
report 




Last 

INCES 

Request 


t 




Determine next module 




* 


t 






RETURN 





Figure 54 : INCES Module Logic 



199 



SAM Supervisor 



I 



Initialize Arrays 



Read Control Card 
Type 90 




Statistical File 



Creation Option 









Read Type 91 Card 


i 


k 




if .present 










Perform 
File 


t File /\ 


Edit 
or . 
Inde 


ing 
xing 


Ma nag 
Appli 


ement ^\. 
cation 





Read Identi- 
fication, 
Link and 
Network MOE 
Data from 
card types 
92->96 



Copy new file 
to statistical 
data tape 






No Type 91 Card 
No comparison to 



be performed 



Locate files to be 
compared and Call 
READCL to prime 
MOE arrays 



Call PRINT to 
report contents 
of arrays 



Call STAT 
to perform 
analysis 



Set Case 
Over Flag 



RETURN 



Figure 55 : SAM Module Logical Flow 



200 



Appendix B 



201 



The PITT Car-Following Model 

1. Symbols Used 

All dimensions are feet and/or seconds 

k = car following parameter (driver sensitivity) 

L = length of the leading vehicle 

T = time scanning interval 

c = lag (the reaction time which is always < T) 

e = maximum emergency deceleration 

x = position of leader at time t 

u = speed of leader at time t 

y = position of follower at time t 

v = speed of follower at time t 

a = acceleration of follower in the interval (t,t+T) 

x* = position of leader at time t+T 

u* = speed of leader at time t+T 

y* = position of follower at time t+T 

v* = speed of follower at time t+T 

b = constant 

2. The Car-Following Model 

The basic assumption is that the following vehicle will 
try to maintain a space headway equal to 

L+10+kv+bk(u-v) 2 (1) 

For the current calculation for the follower we are 
given x*, u*, y, v and we must calculate a. The desired 
position at time t+T is given by (1) as 



x 



*-y* = L+10+kv*+bk(u*-v*) 2 (2) 



but y* = y+vT+aT 2 /2 and v* ■= v+aT and thus expression (2) 
becomes 

x*-(y+vT+aT 2 /2)=L+10+k(v+aT)+bk(u*-v) 2 (3) 



202 



(Note: Since the term (u-v) 2 is sman we have used the 
approximation v* = v. Any difference is taken care of by 
the calibration of b. The increased calculation required 
to use v* is not warranted.) 

Expression (3) gives 

a = 2[x*-y-L-10-v(k+T)-bk(u*-v) 2 ] ,/[T 2 + 2kT] (4) 

This is the basic car following relationship. The 
term involving constant b was introduced to allow for 
high relative closing speed behavior observed empirically. 
The value of b has been calibrated to 



b = 



0.10 for (u-v) < 10 

(5) 
for (u-v) > 10 



The driver reaction time c is introduced into the car 
following equations, after a has been calculated, when the 
new speed and position are defined 

v* = v+a(T-c) 

and y* = y+vT+a (T-c) 2 /2 

where c < T 

3. Derivation of the Emergency Constraint 

This constraint overrides the car following rules to 
prevent collisions. The basic concept provides that the 
following vehicle can stop safely behind its leading 
vehicle under the following conditions: 

(1) The leader decelerates to a stop at the 
maximum emergency deceleration. 

(2) The follower starting at the lag time c later 
decelerates to a stop behind the leader at a 
deceleration rate within the maximum emergency 
deceleration limit. 



203 



If the leader stops at maximum deceleration then 
u* = and 

x* = x+u 2 /2e (6) 

The follower also stopping at maximum deceleration 
will give 

y* = y+cv+v 2 /2e (7) 

Since the headway between the vehicles must exceed 
the length of the leader, expressions (6) and (7) give 

x *_y* = x -y+ (u 2 -v 2 )/2e-cv>L 

and therefore, in general, we have 

x-y>L+cv+ (v 2 -u 2 )/2e (8) 

but x-y>L for all u, v and thus expression (8) holds only 
if 

cv+(v 2 -u 2 )/2e > 

or 

v> (u 2 +e 2 c 2 ) 1/2 -ec 

The basic headway constraint is therefore 

1/2 

x-y>L+cv+ (v 2 -u 2 )/2e if v>(u 2 +e 2 c 2 ) -ec 

(9) 



1/2 



and x-y>L if v < (u 2 +e 2 c 2 ) / -ec 

Suppose that x*, u* , y, v, T are given, then the 
acceleration a of the follower for the time period (t, t+T) 
must be determined such that the headway constraint is not 
violated. 

Two possible cases can arise. 

1. The follower has a speed v* > at time t+T 



204 



2. The follower comes to a stop during the interval 
(t, t+T) . Let this occur at time t+pT where 
< p < 1. 



Case 1: v* = v+a(T-c) 

y * = y+vT+a (T-c) 2 /2 

Case 2: v* = v+a(pT-c)=0 
y* = y-v 2 /2a 

Substituting now for v* , y* in expression (9) 

1/2 
x *-y* > l+cv*+ (v* 2 -u* 2 )/2e if v* > (u* 2 +e 2 c 2 ) -ec 



(10) 



(11) 



(12) 



x*-y* > L if v* < (u* 2 +e 2 c 2 )V 2 -ec 

and from expressions (10), (11) and (12) 
x*-y-vT-a(T-c) 2 /2 > L+cv+ca (T-c) + [ ( {v+a (T-c) } 2 - u* 2 ]/2e 

when v*> (u* 2 +e 2 c 2 )V 2 -ec > 
and x*-y-vT-a (T-c) 2 /2 > L 

when < v* < (u* 2 +e 2 c 2 ) x / 2 -ec 

and x*-y+v 2 /2a > L 

when v* = 
or 

-a 2 [(T-c) 2 /2e]-a[ (T-c) 2 /2 + c(T-c) + 2v(T-c)/2e] 

+ [x*-y-vT-L-cv- (v-u* 2 /2e] > 
when 

v* > (u* 2 + e 2 c 2 ) 1 / 2 -ec > (13) 



205 



and -a[(T-c) 2 /2] + x*-y-vT-L > (13) 

when < v* < (u* 2 +e 2 c 2 ) V 2 -ec (14) 

and a < -v 2 /2 (x*-y-L) 

when v*=0 (15) 

Expression (13) reduces to 

a 2 +a[e+2ec/(T-c)+2v/(T-c) ]- (2e/(T-c) 2 ) [x*-y- 
vT-L-cv-(v 2 -u* 2 )/2e] > 

and this gives 

a < -B/2 + [B 2 +4C] 1 / 2 /2 (16) 

where B = e+2 (ec+v)/ (T-c) 

and C = (2e/(T-c) 2 } [x*-y-vT-L-cv-(v 2 -y* 2 )/2e] 

The condition v* > (u* 2 +e 2 c 2 ) v 2 -ec > reduces to 

v+a(T-c) > (u* 2 e 2 c 2 ) 1 / 2 -ec > 

or 

a > [(u* 2 +e 2 c 2 ) l/2 -ec-v]/(T-c) > (17) 

Expression (14) recuces to 
a < 2 [x*-y-vT-L]/(T-c) 2 
provided 0<v+a(T-c) < (u* 2 +e 2 c 2 ) 1 / 2 -ec (18) 



or 



-v/(T-c) < a < [(u* 2 +e 2 c 2 ) 1 / 2 -ec-v]/(T-c) 



206 



Expression (15) reduces to 

a < -v 2 /2 (x*-y-L) (19) 

provided a < -v/(T-c) 



Expressions (16), (17), (18), and (19) are the con- 
straints which determine the following vehicle's accelera- 
tion which must be maintained to satisfy the emergency 
non-collision conditions. 

Provided the vehicles are in a safe position at time 
t, then the above constaint set will be sufficient for the 
vehicles at time t+T. In particular, B 2 +4C is always 
positive and thus the acceleration given by expression (16) 
has a real value. 

The emergency constraint, however, is also used in the 
lane changing mechanism where the vehicles (in adjacent 
lanes) may not be in a safe position relative to each other 
in a longitudinal sense. In this case the following can 
occur: 

(1) The above constraint set provides a real 
acceleration but it is < e and thus the lane 
change is not initiated. 

(2) The discriminant (B 2 +4C) is negative. In 
this case the lane change is automatically 
not initiated, since the two vehicles must 
be in an unsafe relative position for 
occupying the same lane. 

(3) In the case u* = and x*-y < L, then 
expression (19) operates and gives a spurious 
result. Thus expression (19) is modified 
for lane changing such that the lane change 
cannot be initiated if 

v*=0 and x*-y-L < (20) 

Once again this occurs only if the two vehicles 
are in an unsafe relative position for 
occupying the same lane. 



207 



Appendix C 



208 



ANNOTATED BIBLIOGRAPHY 



1. Car-Following Models 

Bexelius Sten, "An Extended Model for Car Following", 
Transportation Research , Vol. 2, No. 1, March 1968. 

This paper attempts to refine the basic car following 
models by assuming that every driver reacts to several of 
the preceding vehicles. A model is suggested and discussed 
and a stability criterion is found. It is pointed out that 
the "reciprocal-spacing-model" gives a critical velocity 
below which flow is unstable. It is assumed that to load 
a road section to its theoretical capacity, the corres- 
ponding optimum velocity must be in the stable range. 
Based on this assumption, maximum flow is limited by S/2T, 
where S is a constant depending on the ratio of sensitivity 
coefficients and T is response time. 

D. E. Blumenfeld and G. H. Weiss, "Gap Stability in the 
Light of Car Following Theory", Transportation Research , 
Vol. 7, No. 2, June 197 3. 

This article presents an elaboration upon a basic car- 
following model with acceleration noise for predicting the 
statistical properties of a gap measured at two points in 
a traffic stream. If the noise term is assumed to be 
Gaussian, then, in a linear theory, the gap itself will 
have a Gaussian distribution. The variance can be calcu- 
lated, and the resulting theoretical description reproduces 
some of the features found earlier in experimental measure- 
ments by J. H. Buhr. 

A. D. May and E. Keller, "Non-Integer Car-Following Models"* , 
Highway Research Record No. 199 , 1967. 

This paper consists of four major parts. First, a brief 
background is given of microscopic and macroscopic 
theories of traffic flow, with special emphasis on their 
interrelationships. Second, a comprehensive matrix is 
developed which results in the set of steady-state flow 
equations, which includes the major macroscopic and micro- 
scopic theories. Third, analytical techniques are de- 



209 



veloped for evaluating the various theories on the basis 
of experimental data. The last section deals with the 
investigation of a continuum of non-integer car-following 
models for the development of deterministic flow models 
describing the inter-relationships between flow character- 
istics. 

G. F. Newell, "Nonlinear Effects in the Dynamics of Car- 
Following", Operations Research , Vol. 2, No. 9, 1961. 

The purpose of this paper is to show that with a non- 
linear car-following model of the type, V-j(t)=Gj 
[X.;_i (t-A) -Xj (t-A) ] , it is possible to incorporate into a 
single theory all the results previously derived for linear 
car-following models and the nonlinear phenomena previously 
obtained from continuum theories. Thus, the model includes 
most that has been contained in prior models for dense 
traffic flow. In addition, it allows the investigation 
of such things as the development of shocks, shock profiles, 
the range of validity of previous theories and the 
spreading of an acceleration wave. 

It is assumed in this model that the velocity of a car at 
time, t, is some nonlinear function of the space head- 
way at time, t-A, so the equations of motion for a se- 
quence of cars consists of a set of differential difference 
equations. The author argues that there is a special 
family of velocity-headway relations that agrees well with 
experimental data for steady flow and that also gives 
differential equations which can be solved explicitly for 
A=0. Exact solutions of these latter equations show that 
a small amplitude disturbance propagates through a series 
of cars in the manner described by the various theories. 

T. H. Rockewll, R. L. Ernst and A. Hanken, "Sensitivity 
Analysis of Emprically Derived Car-Following Models", 
Transportation Research , Vol. 2, No. 4, 1968. 

This effort examines the sensitivity of empirically de- 
rived regression models of car-following. Data were 
collected from coupled two-car units (a lead car and a 
following car) in light traffic and in dense expressway 
traffic. Inter-vehicular spacing and velocities and accele- 
rations for each of the cars were obtained. The sensi- 
tivities of these variables to the operating conditions 



210 



were modelled. Delays in the variables were investigated 
using a computer simulation with empirical data. The 
principal findings were (1) overall, car-following is 
quite linear, (2) the goodness of fit of any model is 
influenced by stream velocity and time delays. These 
results are discussed in terms of the generality of car- 
following models with respect to the driving population. 

W. E. Wilheln and J. W. Schmidt, "Review of Car-Following 
Theory" , Transportation Engineering Journal of the 
American Society of Civil Engineers , Vol. 99, No. T.E.4, 
1973. 

The authors indicate here that the steadily increasing 
volumes of traffic and the accompanying concern for 
safety have spawned the need for a thorough understanding 
of the dynamic characteristics of vehicular flow. A 
number of theories, approximately 30 models in all, have 
been advanced in the attempt to provide such a mathematical 
description of highway traffic. 

The appeal of this article is the brief history of car- 
following models and a rather comprehensive bibliography 
of closely associated work. 

2. Traffic Simulation Models 

J. H. Buhr, T. C. Meserole and D. Drew, "A Digital Simula- 
tion Program of a Section of Freeway with Entrance and 
Exit Ramps", Highway Research Record No. 250 , 196 8. 

This paper describes a computer program developed for the 
simulation of a section of freeway, including several exit 
and entrance ramps. The program allows for the simulation 
of the traffic operation under different modes of entrance 
ramp control. These are fixed-rate metering, demand- 
capacity metering, gap-acceptance control and no control. 
The computer logic and simulation techniques are discussed 
in detail. Limited output of the program is presented as 
evidence of the feasibility and realism of the simulation 
model. 

Connecticut Department of Transportation, "Traffic Flow 
Simulation Model", 1969. 

This work represents a progress report on a flow simulation 

211 



model which was being developed for highway engineers as a 
tool for investigating, evaluating, and solving expressway 
design problems. The model simulates vehicle flow on 
freeway-type facilities, including all elements of design 
and operation that affect flow and capacity. The effect 
of grades and curvature, sign spacing, and lamp placement 
are included. The model is based on the premise that each 
vehicle in the traffic stream has a desired speed that moti- 
vates its actions. The UNIVAC Model is capable of simu- 
lating a five-mile section of expressway, seven lanes 
directional, with ten on-ramps and ten of f-ramps. The 
vehicle characteristics which seem to have significant 
effect on the model are: gap acceptance, desired speed, 
headway preference, and acceleration and deceleration. 
The model was calibrated by varying these input components 
until the results fell within reasonable statistical range 
of duplicating actual field observations. 

M. J. Craft and J. L. Smith, "Road Traffic Simulation", 
Plessey Communication Journal , Vol. 1, No. 2, 1967. 

This article reviews simulation work in the road traffic 
field and discusses a computer program which has been 
written to simulate traffic passing through a sequence of 
linked traffic signals. By adjustment of the input para- 
meters, a wide variety of traffic situations can be 
studied. The timing of the signals, the number of phases 
and the corresponding traffic movements, the volume of 
through traffic and of traffic turning into and out of the 
intersections, can all be specified by the user, as can 
the speed distributions and other aspects of traffic be- 
havior. The output gives the traffic flows achieved and 
the delays incurred at each stop line. The method was 
tested by comparison with actual observations of a real 
life situation, and it is shown how the program may be used 
to compare the merits of various signal timings, and to 
study the effect of changes in traffic behavior. 

L. Eisenberg and E. Kaplan, "An Investigation of Car- 
Following Model Using Continuous System Model Program 
(CSMP) Techniques", Pennsylvania University Transportation 
Studies Center, Project No. Urt-8, June 1971. 

Quantitative methods for modeling car-following dynamics 
are explored. The behavior of grouped road vehicles is 

212 



predicted according to empirical relationships between 
vehicle performance specifications, roadway surface cond- 
ditions and driver characteristics. By development of 
mathematical equations to correlate these variables, 
traffic dynamics can be simulated without field experiments 
The concept of traffic simulation is described in detail 
along with various stages of model building. The basis 
of the car-following simulation is the continuous system 
modeling program (CSMP) , a special computer program de- 
veloped to integrate various hypothetical conditions. 
Computer printouts are contained in the paper which reveal 
the effectiveness of the CSMP in actual operation. It is 
concluded that traffic flow may be better understood from 
the viewpoint of dynamic simulation modeling. 

Fox and G. Lehman, "A Digital Simulation of Car Following 
and Overtaking", Highway Research Record No. 199 , 1967. 

This article introduces a model representing the single 
lane no-passing driving situation and was formulated and 
run on a digital computer. Although the model involves 
the use of conventional car-following equation, the simu- 
lation also includes human factors, such as reaction time 
lag, driver sensitivity, and the threshold of detection of 
relative velocity. The model allows for variation of 
these characteristics both between drivers and overtime 
for each individual driver. The study was directed to the 
accident prevention problem with the aim of determining 
the critical parameters of the driving situation, and of 
ascertaining the ranges of values of these parameters 
which define a safe or stable driving situation. 

G. W. Harju, "An Advanced Computer Concept for Freeway 
Traffic Flow Modeling", presented at the Summer Simulation 
Conference in San Diego, Calif., June 1972. 

This work introduces a freeway network traffic flow simu- 
lation system developed by the Aerospace Corporation which 
can accept any freeway road configuration under any 
possible traffic conditions. Complex freeway interchanges, 
lengthy networks with any number of on-ramps and off- 
ramps, and intricate weaving sections connecting on- 
ramps and off-ramps can be handled by this program. The 
input module was designed to accept freeway geometry 
identical to the conventions used on the freeway designer's 
coordinate control maps. A unique feature of the model is 



213 



the optional capability to produce computer generated 
traffic flow movies of any specific subarea of the network 
under simulation. The movies appear as stationary over- 
head aerial shots and are used for detailed flow analysis, 
program debugging, and as an aid during the computer 
program validation process. Other features include any 
type of traffic control system, and grade and curve effects 
The simulation is microscopic with random assignment of 
individual driver attributes. Detailed programming has 
produced optimum execution times and the size of the 
freeway network being simulated is restricted only by the 
computer capacity. 

D. R. Korbett, "Digital Simulation Model of Freeway 
Traffic Volume 1: Model Description", Final Report pre- 
pared for the U.S. Department of Transportation, Federal 
Highway Administration, Bureau of Public Roads, Contract 
No. CPR-11-3661, January 1966-November 1968. 

This report describes the freeway simulation model de- 
veloped at the Midwest Research Institute over a number of 
years. A typical general logic is used with vehicles 
attempting to travel at their desired speeds and changing 
lanes to avoid slower vehicles or enter or leave the free- 
way. The car-following algorithm is a fail-safe type 
formulation which is rather complex and includes several 
parameters to be calibrated. The lane changing mechanism 
operates on the decelerations that are required of the 
changing vehicle and its new follower. If these decelera- 
tions are below a certain threshold (termed the acceptable 
risk) , the lane change proceeds with the vehicle changing 
lanes over one time period (one second) . For vehicles 
making forced changes, a capability to scan several po- 
tential gaps to change into is included. The simulation 
includes variable driver and vehicle characteristics and 
has been running successfully although with a rather long 
computer running time. 

U. Larsson and R. Ludin, Urban Traffic Simulation , 
Gothenburg Studies in Business Administration, Basagatan 3, 
Gotenborg, 19 71. 

The stated purpose of the book is first to sketch a 
systems approach to be applied in studies of urban traffic 
systems. Second, to supply a "philosophy" for simulation 

214 



model formulation, and third to present a digital simula- 
tion model to use for the determination of signal programs 
for traffic signal systems in an urban environment. The 
basic version of the model is applicable to pretimed signal 
systems while the expanded version, formulated by Ake 
Lindstrom, is applicable for traffic actuated control. 

Topics included in this book include a discussion of a 
systems approach to urban traffic planning and control. 
This involves the development of a hierarchy of urban 
traffic systems, various control techniques, strategies 
vs. tactics and a working definition of an urban traffic 
system. 

Also included is a brief history and the presentation of 
characteristics associated with traffic simulation models, 
such as actual traffic situations, time-keeping, repre- 
senting the system and the Fortran computer oriented 
characteristics . 

Finally, the basic model is described in terms of modules 
and elements; the use of code numbers, and the use of 
matrices. Representation schemes are given for the control 
system, the vehicles and various events. Some attention is 
devoted to submodels such as for queue discharges, vehicle 
generation and link passage, in addition to a discussion 
of efficiency criteria. 

S. D. Leland, "A General Traffic Flow Simulation Model for 
Freeway Operation", Connecticut Department of Transporta- 
tion in 1970 and by the U.S. Bureau of Public Roads in 
1969. 

A freeway traffic simulation model is presented together 
with a user's manual. The major portion of the paper is 
devoted to the validation effort. The principal goal of 
the project was to develop a model capable of simulating 
vehicle flow on freeway facilities including elements of 
design and operation that affect vehicle flow and capacity. 
The logic of the resultant model is discussed in detail 
with typical values given for the various driver charac- 
teristics that are required as input. Model implementation 
is illustrated with simulation of two freeway sections. 
One, a weaving section with several design alternatives, 
and two, a complex mile section of urban freeway with five 

215 



off-ramps and multiple weaving sections. The model capa- 
city is a five-mile section of expressway with up to seven 
lanes directionally , ten on-ramps , and ten off-ramps with 
up to 1000 vehicles in the system at any one time. The 
program takes up to three minutes to simulate one minute 
of real time. The vehicle characteristics considered 
are: gap acceptance, desired speed, headway preference, 
acceleration and deceleration. The principal limitation 
of the model appears to be that it does not relate vehicle 
performance to the profile of the roadway. 

E. B. Lieberman, "Simulation of Corridor Traffic: The 
SCOT Model", Highway Research Record No. 409 , 19 72. 

The increasing activity in controlling access to freeways 
for the purpose of improving traffic flow has focused on 
the need to develop control policies for treating the 
entire corridor network system. This system comprises 
freeway, servicing ramps, frontage road, and parallel and 
feeder arterials. It has been observed that ramp metering, 
while improving conditions on the freeway, can precipitate 
congestion on the grade roadways. The SCOT model was 
developed as an evaluative and design tool to predict the 
performance of alternative control policies and freeway 
configurations prior to field implementation. A dynamic 
representation of traffic flow is produced by the model. 
This paper describes the capabilities and prominent features 
of the model and some of its representative results. 

E. B. Lieberman, R. D. Worrall and J. M. Bruggeman, 
"Logical Design and Demonstration of UTCS-1 Network 
Simulation Model", 19 72. 

A description is given of a microscopic simulation model 
designed as an evaluative tool for urban traffic control 
policies. The need for such a tool is explored, and the 
underlying benefits of the simulation approach are dis- 
cussed. The logical structure of the model is described; 
the input requirements and statistical output generated 
by this FORTRAN-coded program are detailed, and samples 
are illustrated. 

The requirements and objectives originally assigned to the 
UTCS-1 model are presented. The design of this model for 
the study of urban traffic flow and dynamic signal control 
systems was directed to satisfy the following objectives: 

216 



The model must accurately describe the real-world dynamics 
of urban traffic and respond to a wide variety of controls, 
including responsive systems actuated by an on-line digi- 
tal computer; the traffic dynamics must be expressed in 
terms of significant traffic parameters and measures of 
effectiveness that characterize the performance of each 
component of the network; the accuracy and reliability of 
these results must satisfy the basic research objective 
of utilizing the model as a diagnostic engineering tool 
for the evaluation of alternative policies of traffic 
control, channelizations of traffic, and turning and 
parking restrictions, for any urban network configuration 
and composition of traffic, including buses. 

The resulting model is a balance between considerations of 
sufficient detail to accurately describe the system and 
considerations of practical utility, such as running time 
and core storage restrictions of available computers. It 
is programmed to be acceptable to all computers regardless 
of manufacturer, and it requires a minimum of data acquisi- 
tion and preparation. 

J. H. Mathewson, D. L. Trautman and D. L. Gerlough, "Study 
of Traffic Flow by Simulation". This article appeared in 
Highway Research Proceedings No. 34, dated 1955. 

The paper embraces certain philosophies and approaches in 
the utilization of modern computers to solve traffic prob- 
lems. It concludes that computers, used as simulators, 
offer considerable promise in the solution of such traffic 
problems as investigating the effects of traffic control 
devices in advance of installation and predicting the 
effect of proposed changes on the capacity of a facility. 

The concept of vehicle flow rate or, alternately distri- 
bution of gaps, finds general utility in approaches of both 
analysis and simulation. Ideally, the behavior of a simu- 
lator resembles that of the real situation under study by 
virtue of the postulates laid down by the investigator. 
The authors state that such a model encompasses both the 
structure and the dynamics of the movement of intersecting 
streams of vehicles in terms of flow paths, queueing, 
waiting, proceeding ahead and turning subject to delays 
caused by cross traffic and pedestrians. 

217 



Since this article represents one of the pioneering efforts 
in this area of research it focuses closely on the probable 
types of flow diagrams or algorithms which appeared 
feasible. The study addressed the problem as a "discrete- 
variable simulator" and a "continuous variable model". 

K. C. Sinha and F. Dawson, "Digital Computer Simulation of 
Freeway Traffic Flow", Traffic Quarterly , Vol. 24, No. 2, 
1970. 

It is shown that the techniques of dynamic simulation on a 
digital computer can be well utilized in the analysis of 
traffic flow on multi-lane freeways. A non-technical re- 
view of the most prominent simulation models developed up 
to 1970 is contained in this work, as well as a discussion 
of the essential aspects of such models. 

A comprehensive general model was developed and validated 
to simulate traffic flow on a freeway system with five 
through lanes, four on-ramps and six off-ramps. Presented 
in rather gross mathematical form, the model is intended 
somewhat as a guidepost and can be used to study traffic 
flow on a section of a freeway with a maximum length of 
3 1/2 miles, with both right- and left-hand on-ramps. 

P. Warnshuis, "Simulation of Two-Way Traffic on an Isolated 
Two-Lane Road", Transportation Research , Vol. 1, No. 1, 
1967. 

The author states that one of the open problems in traffic 
flow theory is to describe the flow of two-way traffic on 
a two-lane road. The problem offers many natural complex- 
ities such as hills, curves, intersections and speed 
zones. He indicates that although such factors are 
important from an applied standpoint, they tend to obscure 
the intriguing aspect of the problem: the manner in which 
the interaction between the two lanes affects the flow in 
each. As an aid to developing a theoretical description of 
this interaction, he has constructed a computer simulation 
in which the behavior of individual cars is modelled 
directly. The purpose of the paper is to describe this 
simulation and to present some numerical results obtained 
with it. A good basic list of terms are defined along 
with a clear statement of model assumptions. Model inputs 
are given in non-technical terms as well as ten fundamental 

218 



rules governing simulated flow. 

R. D. Worrall and A. G. R. Bullen, "Lane-Changing on Multi- 
lane Highways", Bureau of Public Roads, Federal Highway 
Administration, U.S. Department of Transportation, Final 
Report, Contract CPR 11-5228, 1969. 

Although this work is quite comprehensive and relates to 
several modeling efforts. Chapter 4, entitled "A Simula- 
tion Model of Lane-Changing on a Multilane Highway" proves 
to be of particular importance to this report. 

This chapter describes a digital simulation model of lane- 
changing behavior. The model simulates the motion of 
individual vehicles in a multilane traffic stream, subject 
to varying assumptions concerning speed and headway dis- 
tributions, car- following rules and gap acceptance. The 
model generates as output counts of lane changes and es- 
timates of lane-change delay, both expressed as functions 
of the input parameters. 



219 



BIBLIOGRAPHY 



Ashton, W. D. , "The Theory of Road Traffic Flow", Kings 
College, London. London: Methuen and Co. Ltd., New 
York: John Wiley & Sons Inc., 1966, p. 21-35, 135-143. 
(Follow-the-leader models of traffic flow- -mathematical 
equations — simulation of traffic problems.) 

Bexelius, S., "Extended Model for Car Following", Trans . 
Res . , Vol 2, No. 1, March 1968, p. 13-21. 

Blumenfeld, D. E. and G. H. Weiss, "Gap Stability in the 
Light of Car-Following Theory", Trans. Res . , Vol. 7, No. 2, 
June 1973, p. 199-205. 

Blunden, W. R. , "The Land-Use/Transport System; Analysis and 
Synthesis", Oxford, N. Y. , Pergamon Press, 1971. 

Breiman, L. Gafarian, A. V., Lawrence, R. and Murthy, V. K. , 
"An Experimental Analysis of Single-Lane Time Headways in 
Freely Flowing Traffic", 4th International Traffic Flow 
Symposium, Karlsruhe, 1968. 

Buhr, J. H., Meserole, T. C. and Drew, D. R. , "A Digital 
Simulation Program of a Section of Freeway with Entrance 
and Exit Ramps", H.R.B. Record No. 2 30, Washington, D.C., 

1968, p. 15-31. 

Buhr, J. H. , et al, "Simulation of Lane Changing on 
Freeways", paper presented at 44th Annual Meeting, H.R.B. , 
1968. 

Bullen A. G. , Worrall, R. D. and Robertson, S., "A Simula- 
tion Model of Lane Changing on a Multilane Highway", H.R.B. 
Highway Research Record, No. 279, Washington, D. C. , 

1969, p. 161. 

Chandler, R. E. , Herman, R. and Montroll, E. W. , "Traffic 
Dynamics: Studies in Car Following", Operations Research 
No. 6 , 1958, p. 165-184. 

Clinton, J. W. , "Lane Changes on an Urban Freeway", John C. 
Lodge Freeway Traffic Surveillance and Control Research 
Project, Detroit, 1962. 



220 



Connecticut State Highway Department, "Traffic Flow 
Simulation Model", 1969. 

Constantine, T. , Young, A. P., "Traffic Dynamics: Car 
Following Studies", Traf. Engr. & Control , Vol. 8, No. 9, 
Jan. 1967, p. 551-554. 

Craft, M. J. and Smith, J. L. "Road Traffic Simulation", 
Plessey Communication Journal , Vol. 1, No. 3, 1967, p. 114- 
121. 

Darroch, J. N. , and Rothery, R. W. , "Car Following and 
Spectral Analysis". 

Davis, E. , "Traffic Engineering Practice", London, E. & 
F. N. Spon, 1968. 

Drew D. R. , "Traffic Flow Theory and Control", McGraw- 
Hill Series in Transportation, New York, 1968, p. 255- 
286; 330-350. (Simulation Techniques, Linear and Non- 
Linear Car-Following Models) . 

Drew, D. R. , Meserole, T. C. and Buhr, J. H. , "Digital 
Simulation of Freeway Merging Operation", Texas Trans. 
Inst. Report 430-6, Texas A&M University, College Station, 
1967. 

Drew, D. R. , "Miccrscopic Models of Traffic Flow", Traffic 
Engr . , Vol. 36, No. 8, May 196 6. 

Edie, L. C. , "Car Following and Steady State Theory for 
Non-Congested Traffic", Operations Research Society of 
America Journal , Vol. 9, p. 72-75, 1961. 

Edie, L. C. , Herman, R. and Rothery, R. , "International 
Symposium on the Theory of Road Traffic Flow", 3rd ed. , 
American Elsevier Pub. Co., New York, 1967. (Held under 
the auspices of Operations Research Society.) 

Eisenberg, L. , Kaplan, E. , "An Investigation of Car- 
Following Model Using Continuous System Model Program 
(CSMP) Techniques", Pennsylvania University Transportation 
Studies Center, Project No. Urt-8, June 1971. 



221 



Everall, P., "Urban Freeway Surveillance and Control — The 
State-of-the-Art", U.S. DOT/FHWA, 1972. 

Fox, P. and Lehman, F. G. , "A Digital Simulation of Car 
Following and Overtaking", H.R.B., Highway Research 
Record No. 199, Washington, D. C. , 1967, p. 33-41. 

Hobbs, F. D. and Richardson, B. D. , "Traffic Engineering", 
Oxford, New York, Pergamon Press, 1967. 

Gafarian, A. V., Walsh, J. E. , "Methods for Statistical 
Validation of a Simulation Model for Freeway Traffic near 
an On-Ramp", Trans. Res ., Vol. 44, No. 4, December 1970, 
p. 379-384. 

Gazis, D. C. , Herman, R. and Potts, R. B., "Car-Following 
Theory of Steady State Flow", Operations Research No. 7 , 
1959, p. 499-505. 

Gazis, D. C. , Herman, R. and Rothery, R. W. , "Nonlinear 
Follow-the-Leader Models of Traffic Flow", Operations 
Research No. 9 , Vol. 4, July-August, 1961, p. 546-567. 

Gazis, D. C. and Weiss, G. M. , "Density Oscillation between 
Lanes of a Multi-Lane Highway", Operations Research No. 5 , 
Vol. 10, September-October, 1962, p. 658-667. 

Gerlough, D. L. , "Simulation of Freeway Traffic by an Elec- 
tronic Computer", Highway Research Board Proceedings 35, 
1956, p. 543-547. 

Gerlough, D. L. , "Simulation of Freeway Traffic by Digital 
Computers", Proc. Conf. on Increasing High Engineering Pro- 
ductivity, Ga. Inst, of Technology, 1956. 

Gerlough, D. L. and Capelle, D. G. , "An Introduction of 
Traffic Flow Theory", H.R.B. Special Report 79 , Washington, 
D.C., 1964. (Queueing Theory Approaches — p. 49; Simulation 
of Traffic Flow — p. 97) . 

Greenshields, B. D. , "Quality and Theory of Traffic Flow: 
A Symposium", Yale University, Bureau of Highway Traffic, 
New Haven, 1961. 

Haight, F. A., "Mathematical Theories of Traffic Flow", 
New York Academic Press, 1963. (Queueing Theory and 
Probability.. ) 

222 



Harju, G.W. , in assoc. with L.R. Bush, R.G. Kremer and 
H.S. Porjes, "An Advanced Computer Concept for Freeway 
Traffic Flow Modeling", The Aerospace Corp., El Segundo, 
Calif., June 14-16, 1972. 

Herman, R. , et al, "Traffic Dynamics: Analysis of Stability 
in Car Following", Res. Lab., Gen. Motors Corp., Detroit, 
Mich., 1958, Operations Research No. 7 , 1959, p. 86-106. 

Highway Research Board, "Traffic Flow Characteristics 196 3 
and 1964", Highway Research Record No. 72, Washington, D.C., 
1965, p. 19 and 58. (A simulation system for study of traf- 
fic flow behavior, interdependence of certain operational 
characteristics within a moving traffic stream. ) 

Highway Research Board, Comm. on Highway Capacity of HRB 
Special Report 87, "Highway Capacity Manual 1965", 
Washington, D.C., 1966. 

Highway Research Board, "Freeway Characteristics, Opera- 
tions and Accidents", Highway Research Record No. 99, 
Washington, D.C. , 1965. 

Highway Research Board, "Planning and Evaluation of Trans- 
portation Systems: 17 Reports", Highway Research Record 
No. 348, Washington, D.C. 1971. 

Jewell, W.S., "Markov Renewal Models in Traffic Flow 
Theory", presented at 3rd International Symposium on 
Theory of Traffic Flow, N.Y. , 1965. 

Kennedy, N. , Kell, J., and Homburger, W. , "Fundamentals of 
Traffic Engineering", 7th ed. , Berkeley, Calif., 1969. 

Knox, D.W., "Merging and Weaving Operations in Traffic", 
Australian Road Res., 2(2), Dec. 1974. 

Kobett, D.R. and Sommerville, D.I., "A Digital Simulation 
Model of Freeway Traffid', (2 Volumes) FHWA, Final Report of 
Contract CPR-11-3661, 1969. 

Kobett, D.R. et al, "Traffic Simulation for the Design of 
Uniform Service Roads in Mountainous Terrain", 4 Volumes, 
FHWA, Final Report of Contract CPR-11-5093, 1970. 

Larsson, U. and Lundin, R. , "Urban Traffic Simulation", 
Institutionen vid Handelshogskolan i Gotenborg, 1971. 

Lee, J. and Jones, J.H., "Traffic Dynamics — Visual Angle 
Car-Following Models", Traf. Eng. and Control , Vol. 9, 
No. 7, Nov. 1967, p. 348-350. 



223 



Lehoczky, J. P. , "Stochastic Models in Traffic Flow 
Theory: Intersection Control", Traffic Flow Theory , HRB, 
Highway Research Record No. 334, Washington, D.C., 1969. 

Leland, S.D., "A General Traffic Flow Simulation Model 
for Freeway Operation:, Conn. DOT, Bureau of Highways, 
Bureau of Planning and Research, May 1970, p. 18. 

Leland, S.D., "A General Traffic Flow Simulation Model for 
Freeway Operations", Conn. DOT, Bureau of Public Roads, 
U.S. , June 1969. 

Lieberman, E.B., "Dynamic Analysis of Freeway Corridor 
Traffic", ASME, Paper Transaction 70-42, Oct. 1970. 

Lieberman, E.B., "Simulation of Corridor Traffic: The 
SCOT Model", HRB, Highway Research Record No. 409, Washing- 
ton, D.C., 1972, p. 34-45. 

Lieberman, E.B. et al, "Logical Design and Demonstration 
of UTCS-1 Network Simulation Model", HRB, Highway Research 
Record No. 409, Washington, D.C., 1972, p. 46-56. 

Makigami, Y. et al, "Three-Dimensional Representation of 
Traffic Flow", Trans. Science , Vol. 5, No. 3, Aug. 1971, 
p. 303. (All flow characteristics depicted as a 3-dimen- 
sional model or space-time graph.) 

Marshall, R.J., "Traffic Simulation Package Using Oppor- 
tunity Model", National Academy of Sciences — National 
Research Council — HRR, No. 250, 1968, p. 8-17. 

Mathewson, J.H. et al, "Study of Traffic Flow by Simula- 
tion", HRB Proc. 34, 1955, p. 522-530. 

Matson, Smith, W.S. and Hurd, F.W. , "Traffic Engineering", 
McGraw-Hill Book Co., New York, Toronto, London, 1955, 
p. 123. (Elements of the traffic stream, probability dis- 
tribution of the stream. ) 

May, Jr., A.D. and Keller, H.E.M., "Non-Integer Car- 
Following Models", HRB, Highway Research Record No. 199, 
Washington, D.C. 1967, p. 19-32. 

May, Jr., A.D. and Keller, H.E.M., "Evaluation of Single 
and Multi-Regime Traffic Flow Models", 4th International 
Symposium on Theory of Traffic Flow, Karlsruhe, June 1968. 



224 



Newell, G. F., "Non-linear Effects in the Dynamics of Car 
Following", Operations Research , No. 9, Vol. 2, March- 
April, 1961, p. 209-229. 

Pahl, J., "Lane-Change Frequencies in Freeway Traffic 
Flow", H.R.B., Highway Research Record, No. 4 09, Washington, 
D. C, 1972, p. 17-25. 

Payne, H. J., "Models of Freeway Traffic and Control", 
Dept. of Electrical Engineering, Univ. of Southern Calif., 
Los Angeles, Calif. 1970. (Car- following models, p. 8). 

Peat, Marwick, Mitchell & Co. and General Applied Science 
Laboratories, Inc., "Network Flow Simulation for Urban 
Traffic Control System", Report No. FH-11-7462-2 , 1971. 

Peat, Marwick, Mitchell & Co. and KLD Associates, Inc., 
"Network Flow Simulation for Urban Traffic Control 
System — Phase II", Report No. DOT-FH-11-7885-4 , 1973. 

Pignataro, L. J., "Traffic Engineering: Theory and Prac- 
tice", Englewood Cliffs, N. J., Prentice-Hall, 1973. 

Pipes, L. A., "Car-Following Models and the Fundamental 
Diagram of Road Traffic", Trans. Res . , Vol. 1, No. 1, 
May 1967, p. 21-29. 

Rockewell, T. H. , Ernst, R. L. , Hanken, A., "Sensitivity 
Analysis of Empirically Derived Car-Following Models", 
Trans. Res ., Vol. 2, No. 4, Dec. 1968, p. 363-373. 

Roy, U. , "Car-Following Models", C.E. 270, Position 
Paper, 1974. 

Rule, R. G. , Fenton, R. E. , "On the Effects of State Infor- 
mation on Driver-Vehicle Performance in Car Following", 
IEEE Trans. Systems, Man & Cybernetics , Vol. Smc-2, 
No. 5, November 1972, p. 630-637. 

Sakarovitch, M. B., "Simulation Model of the Traffic on 
Highway No. 2 between Paris and LeBourget", Automatisme , 
Vol. 15, No. 5, May 1970, p. 226-240. 

Sinha, K. C. , Dawson, R. F. , "The Development and Valida- 
tion of a Freeway Traffic Simulator", H.R.B., Highway 
Research Record No. 308, 1970, p. 34-47. 



225 



Sinha, K. C, Dawson, R. F. , "Digital Computer Simulation 
of Freeway Traffic Flow", Traffic Quarterly , Vol. 24, 
No. 2, April 1970, p. 279-296. 

St. John, A. D. , "Study of Traffic Phenomena through 
Digital Simulation", Final Report for the Public Health 
Service, Dept. of HEW, Research Grant AC-00106, 1966. 

St. John, A. D. , Kobett, D. W. , Glaug, W. D. , Sommerville, 
D. T. and Harwood, D. , "Freeway Design and Control 
Strategies as Affected by Trucks and Traffic Regulations," 
Vol. 1, Tech. Report FHWA, Final Report of Contract 
FHWA-RD-75-T2, 1975. 

Story, H.E.R., "Simulation of Traffic by Digital Computer", 
Traf. Eng. Contr ., Vol. 11, No. 10, Feb. 1970, p. 464-467. 

Tocher, K. D. , "The Art of Simulation", D. Van Nostrand 
Co., Inc., Princeton, N. J., 1963. 

Trautmen, D. L. , Davis, H. , Heilfron, J., Ho, E. C. , 
Mathewson, J. H. and Rosenbloom, A., "Analysis and. Simula- 
tion of Vehicular Traffic Flow", Res. Report No. 20, Inst, 
of Transp. and Traffic Engineering, Univ. of Calif., 
Dec. 1974. 

Underwood, R. T. , "Traffic Flow and Bunching", J. Aust . 
Road Res . , 1(6), June 1963, p. 8-25. 

Warnshuis, P., "Simulation of Two-Way Traffic on an 
Isolated Two-Lane Road", Trans. Res . , Vol. 1, No. 1, 
1967, p. 75-83. 

Wilhelm, W. E. and Schmidt, J. W. , "Review of Car-Following 
Theory", ASCE Transp. Eng. J . , Vol. 99, No. TE 4, Nov. 
1973, Paper 10165, p. 923-933. 

Worrall, R. D. and Bullen, A. G. R. , "Lane-Changing on 
Multi-Lane Highways", Final Report, Contract CPR 11-5228, 
Bureau of Public Roads, FHWA, U.S. DOT, Aug. 1969 (simu- 
lation chapter only) . 



226 



Worrall, R. D. and Bullen, A. G. R. , "Formulating a Model 
of Multilane Traffic Flow", Traffic Flow Theory, H.R.B. 
Highway Research Record No. 334, Washington, D. C, 1969, 
p. 34-38. 

Worrall, R. D. , Bullen A. G. R. and Gur, Y. , "Lane- 
Changing in Multilane Freeway Traffic", H.R.B. , Highway 
Research Record No. 279, 1969, p. 160. 

Wortham, A. W. and Baker, R. L. , "A Macroscopic Event Scan 
Method of Simulation Traffic Flow in a Network", Traf . 
Engr., Inst. Traffic Engr. , Vol. 39, Nov. 1968, p. 42-45. 

Wicks, D. A., et al , "Traffic Flow Simulation Study: the 
SCOT Model", GASL Tech. Report No. 769, Feb. 1972, 
Contract DOT-TSC-161. 

Wohl, M. and Martin B. V., "Traffic System Analysis", 

New York, McGraw-Hill, 1967. (General Traffic Engineering) 



SUGGESTED TOPIC CLASSIFICATION : 

A. Traffic Flow Theory 

B. Car-Following Models — Math 

C. Lane-Changing Models — Delays — Math 

D. Simulation Techniques 



227 



References 



1. NCHRP Report No. 55, "Research Needs in Highway 
Transportation, 1968. 

2. Crane, H. L. , "Proposed Surveillance and Control 
System for the City of Detroit." 

3. Lieberman, E. B. , et al . , "Logical Design and Demonstra- 
tion of UTCS-1 Network Simulation Model," Highway Research 
Record No. 409, 1972. 

4. Peat, Marwick, Mitchell & Co. and KLD Associates, Inc., 
Network Flow Simulation for Urban Traffic Control System - 
Phase II, Report No. DOT-FH-11-7885 , 1973. 

5. AASHO, "A Policy on Geometric Design of Rural Highways," 
1965. 

6. Institute of Traffic Engineers (ITE) , "Traffic Engineer- 
ing Handbook," 1965. 

7. Glennon, J.C. and Joyner, C. A., "Re-Evaluation of Truck 
Climbing Characteristics for Use in Geometric Design", TTI , 
1969. 

8. St. John, A.D. and Kobett, D.R., "Grade Effects on 
Traffic Flow Stability and Capacity", NCHRP 3-19 Draft Final 
Report, MRI, 1974. 

9. Highway Research Board Bulletin 104, "Vehicle Climbing 
Lanes," 1955. 

10. General Motors Corporation, "GMC Intercity Coaches", 
GMC Truck and Coach, Pontiac, Michigan, 1975. 

11. SAE Handbook Supplement 82, "Truck Ability Prediction 
Procedure - J688", Society of Automotive Engineers, 1965. 

12. Western Highway Institute, "Horsepower Considerations for 
Trucks and Truck Combinations, 1969 Acceleration Tests", 
Supplement No. 1, Research Committee Report No.- 2., 1970. 



228 



13. Everall, P.F., "Social Benefits from Minimum Power-Weight 
Ratios for Goods Vehicles", Road Research Laboratory Report 

LR 291, Ministry of Transport, England, 1969. 

14. Highway Research Board Bulletin 107, "Vehicle Operation 
as Affected by Traffic Control and Highway Type", 1955. 

15. Wicks, D. and Cohen, J., SCOT (Simulation of COrridor 
Traffic) Model , Calibration and Validation, Data Acquisition 
and Reduction, prepared for Department of Transportation, 
TSC, under Contract DOT-TSC-740, KLD Associates, Inc., 1974. 

16. Institute of Traffic Engineers (ITE) , "Transportation 
and Traffic Engineering Handbook", 1976. 

17. May, Jr., A.D., and Keller, H.E.M., "Non-Integer Car- 
Following Models", H.R.B., Highway Research Record No. 199, 
Washington, D.C., (1967), pp. 19-32. 

18. Kobett, D.R. and Sommerville, D.I., "A Digital Simula- 
tion Model of Freeway Traffic" (2 Volumes) FHWA, Final 
Report of Contract CPR-11-3661 (1969). 

19. Worrall, R.D. and Bullen, A.G.R., "Lane-Changing on 
Multi-lane Highways", Final Report: Contract CPR 11-5228. 
Bureau of Public Roads, Federal Highway Administration, 
U.S. Department of Transportation, (August 1969). 

20. Harju, G.W. , in Assoc, with Leon R. Bush, R.G. Kremer, 
and H.S. Porjes, "An Advanced Computer Concept for Free- 
way Traffic Flow Modeling", The Aerospace Corp., El 
Segundo, Calif., June 14-16, 1972. 



229 



"\ 



Tl 
O 

l Si 




D 
>— i CD 

s:< 

H CD 

33 »— 

>0 

CO 13 

* 3 

CD 

QJ 3 



O 
c/i 



H 

rn 
o 



CO 

• 

1 > 

— Cc 



FEDERALLY COORDINATED PROGRAM (FCP) OF HIGHWAY 
RESEARCH AND DEVELOPMENT 



The Offices of Research and Development (R&D) of 
the Federal Highway Administration (FHWA) are 
responsible for a broad program of staff and contract 
research and development and a Federal-aid 
program, conducted by or through the State highway 
transportation agencies, that includes the Highway 
Planning and Research (HP&R) program and the 
National Cooperative Highway Research Program 
(NCHRP) managed by the Transportation Research 
Board. The FCP is a carefully selected group of proj- 
ects that uses research and development resources to 
obtain timely solutions to urgent national highway 
engineering problems.* 

The diagonal double stripe on the cover of this report 
represents a highway and is color-coded to identify 
the FCP category that the report falls under. A red 
stripe is used for category 1, dark blue for category 2, 
light blue for category 3, brown for category 4, gray 
for category 5, green for categories 6 and 7, and an 
orange stripe identifies category 0. 

FCP Category Descriptions 

1. Improved Highway Design and Operation 
for Safety 

Safety R&D addresses problems associated with 
the responsibilities of the FHWA under the 
Highway Safety Act and includes investigation of 
appropriate design standards, roadside hardware, 
signing, and physical and scientific data for the 
formulation of improved safety regulations. 

2. Reduction of Traffic Congestion, and 
Improved Operational Efficiency 

Traffic R&D is concerned with increasing the 
operational efficiency of existing highways by 
advancing technology, by improving designs for 
existing as well as new facilities, and by balancing 
the demand-capacity relationship through traffic 
management techniques such as bus and carpool 
preferential treatment, motorist information, and 
rerouting of traffic. 

3. Environmental Considerations in Highway 
Design, Location, Construction, and Opera- 
tion 

Environmental R&D is directed toward identify- 
ing and evaluating highway elements that affect 



• The complete seven-volume official statement of the FCP is available from 
the National Technical Information Service, Springfield, Va. 22161. Single 
copies of the introductory volume are available without charge from Program 
Analysis (HRD-3), Offices of Research and Development, Federal Highway 
Administration, Washington, D.C. 20590. 



the quality of the human environment. The goals 
are reduction of adverse highway and traffic 
impacts, and protection and enhancement of the 
environment. 

4. Improved Materials Utilization and 
Durability 

Materials R&D is concerned with expanding the 
knowledge and technology of materials properties, 
using available natural materials, improving struc- 
tural foundation materials, recycling highway 
materials, converting industrial wastes into useful 
highway products, developing extender or 
substitute materials for those in short supply, and 
developing more rapid and reliable testing 
procedures. The goals are lower highway con- 
struction costs and extended maintenance-free 
operation. 

5. Improved Design to Reduce Costs, Extend 
Life Expectancy, and Insure Structural 
Safety 

Structural R&D is concerned with furthering the 
latest technological advances in structural and 
hydraulic designs, fabrication processes, and 
construction techniques to provide safe, efficient 
highways at reasonable costs. 

6. Improved Technology for Highway 
Construction 

This category is concerned with the research, 
development, and implementation of highway 
construction technology to increase productivity, 
reduce energy consumption, conserve dwindling 
resources, and reduce costs while improving the 
quality and methods of construction. 

7. Improved Technology for Highway 
Maintenance 

This category addresses problems in preserving 
the Nation's highways and includes activities in 
physical maintenance, traffic services, manage- 
ment, and equipment. The goal is to maximize 
operational efficiency and safety to the traveling 
public while conserving resources. 

0. Other New Studies 

This category, not included in the seven-volume 
official statement of the FCP, is concerned with 
HP&R and NCHRP studies not specifically related 
to FCP projects. These studies involve R&D 
support of other FHWA program office research. 



DOT LIBRARY 




0005bT0fl